Parallel processing apparatus for video compression
View Patent ↗A method of and an apparatus for parallel processing video compression is described herein. A picture is divided into N zones vertically, and Vertical Split Parallelism (VSP) deploys N+1 processes: N M-coding processes for general macroblock coding in N zones and one E-coding process for entropy coding. By dividing a macroblock encoding process into M-coding and E-coding processes, multiple M-coding threads and one E-coding thread are able to progress simultaneously with proper synchronization between threads. Since the workload of M-coding processes is greater than E-coding for each macroblock coding process, two cores are able to be designated to the two M-coding processes. Either of the cores are able to be designated to the E-coding process since the workload of E-coding is light. Therefore, VSP utilizes parallel processing computing power efficiently.
1. An apparatus for compressing video comprising:
a. a first processing core for processing a first macroblock layer coding (M-coding) thread; and
b. a second processing core coupled to the first processing core for processing a second M-coding thread,
wherein an entropy coding (E-coding) thread is processed by an available processing core selected from the first processing core and the second processing core, wherein the E-coding thread is based on an output of the first M-coding thread and the second M-coding thread.
2. The apparatus as claimed in claim 1 wherein the first and second M-coding threads include mode decision, transformation and quantization and the E-coding thread includes entropy coding.
3. The apparatus as claimed in claim 1 wherein the available processing core is selected automatically.
4. The apparatus as claimed in claim 1 wherein the first M-coding thread and the second M-coding thread are processed in parallel.
5. The apparatus as claimed in claim 4 wherein the first M-coding thread and the second M-coding thread are processed simultaneously.
6. The apparatus as claimed in claim 1 wherein the first M-coding thread corresponds to a first vertical zone and the second M-coding thread corresponds to a second vertical zone.
7. The apparatus as claimed in claim 1 wherein the first M-coding thread corresponds to a first split-shifted vertical zone and the second M-coding thread corresponds to a second split-shifted vertical zone.
8. The apparatus as claimed in claim 1 wherein the first M-coding thread corresponds to a first diagonal zone and the second M-coding thread corresponds to a second diagonal zone.
9. A method for compressing video comprising:
a. dividing a picture vertically into N zones, wherein N is greater than or equal to two;
b. deploying N+1 processes, wherein N processes are macroblock layer coding (M-coding) and one process is entropy coding (E-coding);
c. processing the N M-coding processes in parallel utilizing a plurality of processing cores;
d. processing the E-coding process to process an output of the M-coding processes utilizing an available processing core of the plurality of processing cores; and
e. synchronizing the N+1 processes.
10. The method as claimed in claim 9 wherein processing the N M-coding processes includes M-coding a current macroblock after M-coding a previous macroblock finishes.
11. The method as claimed in claim 9 wherein processing the E-coding process includes E-coding the current macroblock after M-coding the current macroblock finishes.
12. The method as claimed in claim 9 wherein processing the E-coding process includes E-coding the current macroblock after E-coding the previous macroblock finishes.
13. The method as claimed in claim 9 further comprising:
a. M-coding a current macroblock after M-coding a previous macroblock finishes;
b. E-coding the current macroblock after M-coding the current macroblock finishes; and
c. E-coding the current macroblock after E-coding the previous macroblock finishes.
14. The method as claimed in claim 9 further comprising M-coding a macroblock after E-coding of an upper-right macroblock finishes if the macroblock is not in a first macroblock row.
15. The method as claimed in claim 9 further comprising M-coding a macroblock after E-coding of an upper macroblock finishes if the macroblock is a last macroblock in a macroblock row.