Alternative Spliterator for SpinedBuffer
peter.levart at gmail.com
Thu Apr 18 08:42:04 PDT 2013
On 04/18/2013 04:39 PM, Peter Levart wrote:
> I did some performance measurements. The most obvious difference in
> performance is observed when individual operations in the stream
> pipeline take relatively long time, but even with fast individual
> operations, the performance benefit can be observed.
Well, not really. The length of individual operations actually does not
matter much, but the size of spliterator. The size of leaf split for
iterative FJ task is determined by the number of elements in
spliterator. If it is larger, the number of sub-optimal tasks from
default spliterator is larger.
More information about the lambda-dev