While the single benchmark provided probably isn't statistically significant, I noticed that performance on system with 4 cores of less is worse with this patch.
If this is actually symptom of the patch, would it be possible to toggle lock behavior depending on available cores?