> I implemented restartable atomic sequences (RAS) to compare their
> performance with atomic test-and-set instructions.  The RAS facility
> is described in
> 
> http://www.cs.cmu.edu/afs/cs/project/mach/public/doc/published/Rcs.ps
> 
> and is required on the MIPS R3000 cpu for libpthread support.  But
> the question arises whether the RAS facility would be more broadly
> useful...

The MI implementation can be found here:

	ftp://ftp.netbsd.org/pub/NetBSD/misc/gmcgarry/

Includes support for mips, m68k and i386.  The man pages discuss
most of the issues.  Would appreciate comments before moving forward.

	-- Gregory McGarry <g.mcgarry@ieee.org>