FIGURE 11.

Execution overhead (in microseconds) of the Op benchmark.increment(data: byte[]), through static method invocation, SciJava Ops, and ImageJ Ops. Op adaptation [A] was additionally benchmarked for SciJava Ops by requesting the Op as a function Op, and for static invocation by creating an output analogous to the Op adaptation routine. Parameter conversion [C] was benchmarked for SciJava Ops by requesting the Op with a double[] parameter, and conversion was performed manually for static invocation for comparison. Finally, a composition of adaptation and conversion [A + C] was benchmarked to profile the overhead of their combination. In all SciJava Ops benchmarks the Op cache was disabled to ensure the performance of the Op matcher was measured, except in the case where caching was explicitly enabled for the sake of comparison to static invocation. Execution overhead is displayed on both a logarithmic (left) and linear (right) scale for comparison.