Efficient index computation for array based structured data

Traditional computing techniques widely vary over large scale computing. Some techniques are useful for structured data while some other techniques are useful for non-structured data. To store and operation on the structured data, multidimensional array is widely used as basic data structure. But the performance of multidimensional array is very poor when the number of dimension is very high i.e. 32-dimension. Data scientist suggests linearization of dimensions for higher dimensional data. But linearized data are highly compute intensive for retrieving into original data format as well as operations are very expensive on that data. It suffers from the higher index computational cost and lower data locality. In this paper, we propose an index computation algorithm based on generalized two-dimensional structure. We showed that the accessing cost of array elements is higher for traditional multidimensional array over generalized two-dimensional representation. Though the most sequential access of main memory is itself very simple for any compiler yet experimental results showed a better performance. Both theoretical analysis and experiment are done to show its' effectiveness of the proposed algorithm.