To model TG and TC, we introduce the total workload
size (W) and the processing throughput (P). Assuming the
application problem size is n, and the workload for each
data point i in the problem space is wi, we define the total
workload size as W ¼ Pn1
i¼0 wi (i.e., the sum of the workload
of all the data points). To simplify the model description,
we assume that each data point performs the same