摘要 |
A computer-implemented method for optimizing at least part of data processing pipeline comprising: determining at least a first data processing unit associated with a first data operation and second data processing unit associated with a second data operation in a data processing pipeline, a data output of said first data processing unit being used as a data input of said second data unit; and, fusing said first and second data processing unit into a third data processing unit associated with the combined operations of said first and second data processing units, said fusing including: patching at least part of a code representing the operation of said first data processing unit into a mutable data part of said second data processing unit, said mutable data part comprising mutable code representing the operation of said second data processing unit; and, combining the patched code of said first data processing unit with the mutable code of said second data processing unit into mutable code representing the combined operations of said first and second data processing units. |