[PATCH] memorder: Reconcile W+RWC propagation path with litmus dataflow
Kunwu Chan <[email protected]> Wed, 11 Mar 2026 15:58:06 +0800
| Newsgroups | org.kernel.vger.perfbook |
|---|---|
| Message-ID | <[email protected]> |
Align the Section 15.3.7.2 propagation narrative with Listing 15.18 by correcting the path to P2->z->P1. This removes a variable-path contradiction in the counter-temporal explanation and restores argument traceability. Signed-off-by: Kunwu Chan <[email protected]> --- memorder/memorder.tex | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/memorder/memorder.tex b/memorder/memorder.tex index 16a569d1..2357e7ea 100644 --- a/memorder/memorder.tex +++ b/memorder/memorder.tex @@ -2755,8 +2755,8 @@ This situation might seem completely counter-intuitive, but keep in mind that the speed of light is finite and computers are of non-zero size. It therefore takes time for the effect of the \co{P2()}'s store to -\co{x} to propagate to \co{P1()}, which in turn means that it is possible -that \co{P1()}'s read from \co{x} happens much later in time, but +\co{z} to propagate to \co{P1()}, which in turn means that it is possible +that \co{P1()}'s read from \co{z} happens much later in time, but nevertheless still sees the old value of zero. This situation is depicted in \cref{fig:memorder:Load-to-Store is Counter-Temporal}: -- 2.25.1