Some degradations arise for 10-byte messages; these are an unrealistic data point in well-tuned applications but may be
useful for considering applications for which the communication is difficult to coordinate or for benchmarking the base latency of a communication channel.
In general, however, the curves indicate better performance benefits from larger messages or larger N values. These results are not surprising given the greater amount of copying (and expense of copying) required in such tests.
At best, the results indicate a 40% reduction in CPU execution time after adding the new interface.