Re: [PATCH v2] tools/memory-model: Add extra ordering for locks and remove it for ordinary release/acquire

[Michael Ellerman]

Michael Ellerman <mpe@xxxxxxxxxxxxxx> writes:
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> writes:
...
>
> I'll do some kernbench runs tomorrow and see if it shows up there.

Finally got some numbers.

The summary is it's in the noise, the average delta is +0.1s (total time
is ~67s, so 0.17%) but the standard deviation is ~0.2s.

So that's probably a real slow down, but it's very minor.

I guess that's not surprising, given kernbench should be spending the
bulk of its time in userspace.

cheers