Runs the convolution operator on the specified input file. The base ConvolutionOperator runs the
identity kernel i.e. input and output data should be identical.
If an accelerator that supports OpenCL is found, the convolution operation is
performed on the "best" (as determined by number of compute units
https://software.intel.com/sites/landingpage/opencl/optimization-guide/Basic_Concepts.htm ) device available.
If no accelerator is found, the convolution operation is performed on the CPU using Java Thread Pools. This is
also the case if the best device is a CPU - currently OpenCL on a CPU greatly underperforms native JTP.