memorder: Wordsmith quick quiz on x86 memory ordering

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>

1 files changed, 5 insertions, 5 deletions

diff --git a/memorder/memorder.tex b/memorder/memorder.tex
index c45ac4f6..a53dd8b5 100644
--- a/memorder/memorder.tex
+++ b/memorder/memorder.tex
@@ -2949,10 +2949,10 @@ passage of time itself as a memory-ordering operation.
 	\cref{fig:memorder:Load-to-Store is Counter-Temporal on x86}
 	dashes any fond hopes that loads executing after a given store
 	would see that store's value (or some later value).
-	This data is generated by a program in which the parent thread
-	spawns the children, each of which polls a shared variable.
-	After all the children are running, the parent writes a new
-	value to the shared variable.
+	This data is generated by 1,000 runs of a program in which the
+	parent thread spawns the children, each of which polls a shared
+	variable.  After all the children are running, the parent writes
+	a new value to the shared variable.
 	The quantity histogrammed in the figure is the time from just
 	after the store until just before the last load of the old value.
 	And most of these data points lie on the negative x~axis, clearly
@@ -2978,7 +2978,7 @@ passage of time itself as a memory-ordering operation.
 	as shown in
 	\cref{fig:memorder:Is Store-to-Load Counter-Temporal on x86?},
 	even on weakly ordered systems.
-	And all the data points in
+	And the data points from all 1,000 runs shown in
 	\cref{fig:memorder:Store-to-Load is Temporal on x86},
 	lie on the positive x~axis,
 	so in this case, our temporal hopes have been fulfilled.