CVS difference for ai12s/ai12-0251-1.txt
--- ai12s/ai12-0251-1.txt 2018/11/22 02:11:40 1.6
+++ ai12s/ai12-0251-1.txt 2019/01/25 23:25:51 1.7
@@ -1,4 +1,4 @@
-!standard 5.5(3/3) 18-11-16 AI12-0251-1/03
+!standard 5.5(3/3) 19-01-25 AI12-0251-1/04
@@ -12,7 +12,7 @@
!subject Explicit chunk definition for parallel loops
-The iterations of a parallel loop are grouped into "chunks," optionally with
+The iterations of a parallel loop are grouped into "chunks", optionally with
a chunk parameter identifying the chunk associated with a given iteration.
This supports a roll-your-own parallel map/reduce, as well as providing
control over the level of parallelism to use to execute a given parallel loop.
@@ -26,8 +26,8 @@
However, existing loops are not easy to convert into the reduction attribute
form, especially if the loop has multiple accumulators. They could be written
as multiple reduction expressions, but that could require recalculating parts
-of the expression multiple times, harming performance (which is the point
-of parallelizing in the first place). They also could be written using a
+of the expression multiple times thereby harming performance (which is the
+point of parallelizing in the first place). They also could be written using
a protected object to model the accumulators, but that would force some
serialization of the loops (again, harming performance).
Questions? Ask the ACAA Technical Agent