RE: Poor thread performance on Linux vs. Solaris

> From: Bill Soudan [mailto:[email protected]]


> 4) try to eliminate the global locks in the kernel/futex.c file.  I'm
> afraid it might not be that easy though - the weird freeze make me wonder
> if the scheduler may play a role here as well.

I'd bet my neck on this one -- the freeze could be the scheduler
considering the spinning being CPU time burned and thus being taken
off the run list until all other tasks expire their timeslices.

A coworker of mine played with the notion of removing global lock
for the futexes, replacing it with a per-hash-queue lock. He did it
on our rtfutex patch, and I don't know how easy would it be for 
the mainstream futexes to adapt, but it might be worth a try; it
will not resolve fully the problem, but it might help a lot.

> 5) make some noise on linux-kernel.  Figured I'd check here first.

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