- ...constructs
- We have freely used N-ary in the code
fragments to stand for unary, binary, ternary, 2#2. The actual
system has separate constructs for each of these cases.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...(N-1)-ary
- read
(N-1)-ary as (N minus 1)-ary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...implementations
-
It may seem that making assumptions about
sizes of ints and shorts is against the object-oriented
paradigm, it is important for obtaining good performance. ints
and shorts can be used in conditional expressions and can be
checked using a single scalar compare operator. Since the fields on
which the condition checks are made are prone to shared-memory bottlenecks
such optimizations contribute to significant performance improvements.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...follows
- The declaration shown here uses
template parameter constraint where ReducerType is specified to be of
type binary_function. This is only for documentation purposes. Our
compiler did not support specifying template constraints at the time
of this implementation.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
