RE: Poor condvar performance

> From: Jamie Lokier

> Real-time futexes have been raised on linux-kernel long ago as I'm
> sure you know, but the proposal and patch then implied a particular
> choice of real-time futex wakeup policy that contradicts the FIFO
> wakeup principle which some code assumes.  

Hmmm, I am almosty ready to release some new patches with all this 
stuff, and this sounds interesting. I never heard (my fault probably) 
of a FIFO argument; I'd like to know which are the places where that 
order is assumed to see what could be done.

However, the fact that we are using a wait queue to implement a mutex
is a consequence of this. My take is that the solution steps through
splitting both concepts.

Iñaky Pérez-González -- Not speaking for Intel -- all opinions are my own (and my fault)