Scheduling of turbo decoding on a multiprocessor platform to manage its processing effort variability

This paper presents means of mapping and scheduling portions of a wide-band code division multiple access (WCDMA) application on a homogeneous multi processor system-on-chip (MPSoC). We focus on the turbo decoder, which is a computationally intensive part of the application and which presents a significant processing variability. Our model allows deriving and validating a flexible scheduling method for turbo decoding tasks, which is adapted to the variable processing effort required by the decoder. Using a proposed performance model, the efficiency of this scheduling method is demonstrated. A proposed flexible scheduling (FS) method, when compared to a worst case execution time (WCET) scheduling method, allows increasing the number of users from 14 to 29, while keeping an acceptable quality of service, as reflected in a very small degradation of less than 0.15 dB of the decoding gain.