Subj : Re: atomic operations api added to AppCore...
To   : comp.programming.threads
From : Ronald Landheer-Cieslak
Date : Thu Feb 24 2005 12:47 pm

awaken wrote:
> Can you give me email? I'm not sure is it right place to ask.
> Page 5, StabilizeRight procedure:
> There are some undefined variables (marked with bold font)
> <prevnext, t> <- prev.^Right;
> ....
> if!CAS(&prev.^Right,<prevnext,t>,<r,t+1>) return
> 
> where this "t" came from and what does it mean?
> there is also no definition for prevnext variable, I guess is's something
> like this->prev->next (in terms of C++)
For memory management, he uses an ABA tag and hazard pointers, AFAICT,
you only need either one of the two.
The first time he uses t, on line 6 of the algo of StabilizeRight, I
think it's a type-o, as there would be no reason for Anchor to contain
anything but the status on the third field, which is s, not t.
On line 7, t gets assigned whatever was in the integer field of the
right-hand node, which is basically an anti-ABA tag. Again, if you use
hazard pointers, you don't need that.

My current version uses hazard pointers - so I don't use the ABA tag as
hazard pointers are immune to ABA. If you're using hazard pointers, you
won't need them either..

HTH

rlc

.