Interuniversitair Micorelektronica Centrum (imec, Vzw)

Imec

Interuniversitair Microelektronica Centrum (imec)

Erik Brockmeyer

Francky Catthoor

Tanja Van Achteren

Cedric Ghez

Johan D'Eer

In one aspect, there is a formalized method for mapping applications on a multiprocessor system. In particular re-use possibilities are explored, e.g. focus on data transfer and memory access issues, with the aim of obtaining low-power and low-energy mappings and/or to overcome memory performance or energy bottlenecks.

Method for mapping applications on a multiprocessor platform/system

The invention presents video information stream encoding methods used in applications of a data storage and transfer design methodology for data-dominated applications. The invention relates to video encoding methods with variable video frames designed such that the digital system on which the methods are implemented, consumes a minimal of power, during their execution and still obtain excellent performance such as speed compliance. The resulting video information stream encoding methods can be mapped on different processor architectures and custom hardware. The methods enable combined low power consumption, reduced bus loading and increased performance to achieve speed compliance. The encoding methods are essentially based on block-based motion estimation and grouping of motion estimations of various video frames.

Optimized data transfer and storage architecture for MPEG-4 motion estimation on multi-media processors

System and method of encoding video frames

The power consumption, memory allocation, and CPU time used in a signal processor executing a fast transform can be optimized by using a particular method of scheduling the calculations. The transform typically is used to transform a first m-dimensional indexed array into a second m-dimensional indexed array. The elements of the first m-dimensional array are grouped according to the index difference between the elements of the particular butterfly code of that stage. A second grouping of elements is composed of butterfly code elements having non-maximal index differences. The second group advantageously includes elements also assigned to the first group. The butterfly codes of the groups are arranged sequentially and are executed in the sequential schedule. In a second embodiment, elements are grouped according to a group specific threshold value. Memory is allocated to the groups according to the size of the group such that the access to memory is minimized during execution.

Method for performing a fast transform

Growing community of inventors

Leuven, Belgium

Erik Brockmeyer