Re: [PATCH] Yielding processor resources during lock contention

[Anton Blanchard]

Hi,

> For s390, this was solved by simply defining cpu_relax() to the hypervisor
> yield operation, because we found that cpu_relax() is used only in busy-wait
> situations where it makes sense to continue on another virtual CPU.
> 
> What is the benefit of not always doing a full hypervisor yield when
> you hit cpu_relax()?

cpu_relax doesnt tell us why we are busy looping. In this particular
case we want to pass to the hypervisor which virtual cpu we are waiting
on so the hypervisor make better scheduling decisions.

Did you manage to see any improvement by yielding to the hypervisor
in cpu_relax?

Anton
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/