Is using taskkill bad?

[Stonecold]

If I were to create a batch file that terminated Folding@home using the taskkill command, would that be bad for it? Is there a difference between shutting the core down by exiting the command windows and using something like "taskkill /IM FAH.exe" or "taskkill /IM FahCore_a3.exe"? Because if exiting it by closing the command window allows it to save its work progress and shut down properly, then using the taskkill command on it would be bad (taskkill immediately terminates a process without interacting with it or "asking" it to shut down by itself). Or is there another batch command that would work to safely terminate a core?

[MysticGold04]

I don't think its bad... from time to time I kill my folding process by issuing this command in linux:

Code: Select all

killall fah6

it seems to stop it gracefully and it continues normally once I restart it.

[bruce]

Abnormal termination of cores (including taskkill) increases the chances of loss of data. It might work a lot of the time (as long as you still shut down Windows cleanly) but you will lose data sometimes so I don't recommend it.

In Linux, there are several options for the kill command, not just the unconditional "kill it now" choice that Windows provides with taskkill. I'm not enough of a Windows expert to know if those other options exist somewhere in the Windows CMD interface. The Linux gurus have worked out scripts which terminate a V6 client and its cores with no more damage to the checkpoints than using CTRL-C or logging out. I have not seen a similar script developed for Windows.

Development is currently looking at how the cores get terminated whenever the the V7 client is terminated. The next open beta will behave differently than the current open beta.

If FahCore_78 is terminated rather than allowing it to shut itself down, optimizations will be disabled in the next run.

[Stonecold]

Maybe I should use Linux, then... Does Linux process WUs faster?

[bruce]

Does Linux process WUs faster?

It depends on which FahCore you're asking about and which version is active.

For a while, the Linux SMP core was a later version than the Windows version, and Linux was faster. Since the FahCore versions are now the same, I'm not sure. Performance may very well be identical. I have not seen a head-to-head comparison recently. It probably also depends on whether the core is compiled for 32-bit or 64-bit, especially where there CPU counts in a system are high.

Supported GPU cores run only on Windows (though it's possible to run NVidia GPUs on Linux with wine plus a 3rd party supported interface), so no, Linux IS NOT faster.

For a while Linux had no uniprocessor support but with the advent of FahCore_a4, it's now possible to run a uniprocessor client on Linux as well as Windows. performance is probably identical.