Method for processing an object on a platform having one or more processors and memories, and platform using same
View Patent ↗A method for processing, in a platform having one or more processors and memories, an object including elementary information of same type. The method decomposes the object to be processed into at least two sub-objects of N elementary information each, all the sub-objects having the same quantity N of elementary information, the processing performing at least one sequence of specific operations on the elementary information of each sub-object performing, for each sub-object, at least N times each specific operation, such that each elementary information for each sub-object is applied at least one time, and N results are produced for each specific operation, the sequence of specific operations being such that at least one specific operation of the sequence produces, at least once during its N applications, a result used for the processing of another sub-object.
1. A platform for processing an object having at least two dimensions and including a plurality of pieces of elementary information, each piece of elementary information being represented by at least one numerical value, said platform comprising:
at least one queue and N*P processors, according to at least one circular chaining, located according to a grid of N*P processors corresponding to a rectangle of N processors horizontally and P processors vertically, each processor performing a sequence of specific operations on sub-objects of the object that have a number of pieces of elementary information identical to the number of processors, said specific operations including calculation of an arbitrary number of vertical and/or horizontal and/or non-separable filters;
a transferring unit to transfer a result of at least one same specific operation of the sequence of specific operations performed by each of the processors to a processor or queue that follows according to the circular chaining,
wherein
the queue receives at least one result of said same specific operation performed by one of the processors of the circular chaining located at a first extremity of the grid, and
another processor of the circular chaining located at a second extremity of the grid receives at least one result of the same specific operation of said queue performed during application of the sequence of specific operations to a preceding sub-object by said processor of the circular chaining located at the first extremity of the grid.
2. The platform according to claim 1 , wherein P is equal to 1.
3. The platform according to claim 1 , where the object is an image and a piece of elementary information is a pixel.
4. The platform according to claim 1 , wherein the number of queues is the same as the number of dimensions of the sub-objects.
5. A method, implemented on a platform, for processing an object having at least two dimensions and including a plurality of pieces of elementary information by a platform, each piece of elementary information being represented by at least one numerical value, said platform including, according to at least one circular chaining, at least one queue and N*P processors located according to a grid of N*P processors corresponding to a rectangle of N processors horizontally and P processors vertically, said method comprising:
performing, via each processor, a sequence of specific operations on sub-objects of the object that have a number of pieces of elementary information identical to the number of processors, said specific operations including calculation of an arbitrary number of vertical and/or horizontal and/or non-separable filters on at least one piece of elementary information;
transferring a result of at least one same specific operation of the sequence of specific operations performed by each of the processors to a processor or queue that follows according to the circular chaining;
receiving, in said queue, at least one result of said same specific operation performed by one of the processors of the circular chaining located at a first extremity of the grid; and
receiving, in another processor of the circular chaining located at a second extremity of the grid, at least one result of said same specific operation of said queue performed before by said processor of the circular chaining located at the first extremity of the grid.
6. The method according to claim 5 , wherein P is equal to 1.
7. The method according to claim 5 , wherein the object is an image and a piece of elementary information is a pixel.
8. The method according to one of claims 5 , wherein the number of queues is the same as the number of dimensions of the sub-objects.