Subj : Re: Memory visibility and MS Interlocked instructions
To   : comp.programming.threads
From : Peter Dimov
Date : Sat Sep 03 2005 02:28 pm

Alexander Terekhov wrote:
> Peter Dimov wrote:
> [...]
> > I find that very odd. Not the absence of atomic visibility per se,
> > but...
> >
> > P1: X = 1; #storestore; Y = 1;
> > P2: if( Y == 1) { #loadstore; Z = 1; }
>
> That's dependent store, no barrier needed.

Added for completeness' sake.

> > P3: if( Z == 1) { #loadload; assert( X == 1 ); }
> >
> > If Z == 1 happens before X == 1, Z = 1 happens-before Z == 1, Y == 1
> > happens-before Z = 1, Y = 1 happens-before Y == 1 and X = 1
> > happens-before Y = 1, then X = 1 happens-before X == 1, right?
>
> Right. Under PC,
>
> P1: X = 1; Y = 1;
> P2: if( Y == 1) { Z = 1; }
> P3: if( Z == 1) { assert( X == 1 ); }
>
> Z == 1 && X != 1 is false.

Great. But

P1: Y = 1;
P2: if( Y == 1 ) { Z = 1; }
P3: if( Z == 1 ) { assert( Y == 1 ); }

may assert. This still breaks happens-before; there is a similar Y = 1
<< Y == 1 << Z = 1 << Z == 1 << (other) Y == 1 chain.

.