In order to meet the timeliness requirements of evaluation work, the research on the solution speed of large-scale Data envelopment analysis (DEA) problems has received attention for a long time. Although the Framework method has the fastest computing time in literature, it still has possibility to improve. This paper presents a new parallel computing algorithm work better than the Framework. In this algorithm, the computation speed of large-scale DEA models is improved by combining traditional DEA solving algorithm with one-dimensional parallel hyperline algorithm, and speed up the solution of multi-dimensional problems through dimensionality decomposition. This method does not need to set parameters to solve the problem like the Framework. The advantages of the new proposed algorithm are verified by the results from case studies with randomly generated datasets and real-life case. To ensure the comprehensiveness of the results, randomly generated datasets are under the assumption that obey the uniform distribution and the normal distribution respectively. Very large scale problems, exceeding fifty million decision making units (DMUs), were solved for the first time by our algorithm. The bankruptcy data of Polish companies is used to test the computing ability and availability of our algorithm in real-life case.