Apparatus, method and data processing element for efficient parallel processing of multimedia data

A data processing apparatus for efficient parallel processing of multimedia data in a DPE of a parallel processing system comprising a plurality of DPEs, the apparatus comprising;

• one or more local memory pages comprising instructions and local data;
• a virtual page which is selectively connected to one among shared memory pages in the DPE and adjacent DPEs;
• and
• a dynamic remapper assigning a shared memory page to the virtual page or one of the adjacent DPEs according to conditions that a series of tasks for processing multimedia data are processed in the DPE and the adjacent DPEs.

A method of processing data in a parallel processing system comprising a plurality of DPEs, wherein each DPE comprises one or more local memory pages and a shared memory page shared by the DPE and DPEs that are adjacent to the DPE, the method comprising:

• dividing a multimedia data processing algorithm into at least two sub-blocks that can be processed in parallel;
• assigning a series of tasks for processing the sub-blocks to the DPE and the adjacent DPEs considering the directions at which the adjacent DPEs are disposed with respect to the DPE, according to a performing order of the tasks;
• and
• dynamically remapping the shared memory pages of the DPE and the adjacent DPEs according to conditions that the series of tasks are performed in the DPE and the adjacent DPEs.

A data processing method in a DPE of a parallel processing system when the parallel processing system divides a multimedia data processing algorithm into a series of at least two tasks and performs the tasks in parallel, wherein the parallel processing system comprises a plurality of DPEs, and the DPE comprises one or more local memory pages and a shared memory page shared by the DPE and DPEs that are adjacent to the DPE, the method comprising:

• importing data stored in the shared memory page of the DPE and performing a task assigned to the DPE;
• remapping a shared memory page of a first DPE of the adjacent DPEs so that data generated by the DPE is stored in the shared memory page of the first DPE of the adjacent DPEs, which performs a task subsequent to a task performed by the DPE;
• and
• storing the data generated by the DPE in the remapped shared memory page.