Re: [PATCH] doc: memory-barriers.txt: Correct example for reorderings

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Thu, Aug 28, 2014 at 05:45:10PM -0400, Pranith Kumar wrote:
> On Thu, Aug 28, 2014 at 5:39 PM, Paul E. McKenney
> <paulmck@linux.vnet.ibm.com> wrote:
> > On Thu, Aug 28, 2014 at 05:15:50PM -0400, Pranith Kumar wrote:
> >> On Thu, Aug 28, 2014 at 5:00 PM, Paul E. McKenney
> >> <paulmck@linux.vnet.ibm.com> wrote:
> >> >
> >> > I am really confused by the ordering of the new lines, as I would have
> >> > expected the "x"s to be replaced with "y"s and vice versa.  But the
> >> > ordering of the various combinations does not matter, and each line
> >> > looks OK.
> >>
> >> Since x = A was replaced by x = B and y = B was replaced by y = A, I
> >> just converted all x = load A to x = load B. Similarly for y. I think
> >> we can do either this or as you suggested replacing x's with y's.
> >
> > Fair enough.  The other thing that confused me was the "STORE B=4,"
> > showing up early in your replacement.
> 
> So I converted
> 
> STORE A=3, x=LOAD A->3
> 
> to
> 
> STORE B=4, x=LOAD B->4
> 
> since the later load into x wanted an updated value of B and not the
> default. That is the reason you see STORE B = 4  in place of STORE A =
> 3

Fair enough.

However, the list is supposed to be part of the full set of 24 possible
outcomes.  Many readers would therefore expect the permutations to be
ordered in some way, so that the first column would have six of one thing,
six of the next thing, and so on, according to some pattern.

Could you please adjust in this way?  Again, the point is to be the
first few items in a full set of permutations, not to illustrate selected
orderings/outcomes.

							Thanx, Paul