US 12,333,626 B2
Processor, method and non-transitory computer-readable storage media for handling data
Rune Holm, Oslo (NO); and Elliot Maurice Simon Rosemarine, London (GB)
Assigned to Arm Limited, Cambridge (GB)
Filed by Arm Limited, Cambridge (GB)
Filed on Mar. 15, 2023, as Appl. No. 18/184,212.
Prior Publication US 2024/0311947 A1, Sep. 19, 2024
Int. Cl. G06T 1/20 (2006.01); G06T 1/60 (2006.01)
CPC G06T 1/20 (2013.01) [G06T 1/60 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A processor for handling data, the processor comprising a handling unit, a plurality of storage elements, and a plurality of execution units, the processor configured to:
obtain, from storage, task data describing a task to be executed in the form of a plurality of operations on data, wherein each of the operations maps to a corresponding execution unit of the processor, the task data further defining an operation space of said data, the operation space representing dimensions of a N-dimensional arrangement of the operations to be executed, wherein N>1;
wherein the handling unit is configured to, for each of a plurality of N dimensional portions of the operation space:
define transformation data comprising transformation instructions representing a transform of the N dimensional portion of the operation space to generate respective operation-specific local spaces for each of the plurality of the operations, wherein said transformation data comprising a data field corresponding to each of the N dimensions of the operation space;
for at least one operation of the plurality of operations, upon transformation from the operation space to the operation-specific local space requiring transformation instructions producing M dimensions, wherein M<N, define one or more operation-specific arguments, and store said one or more operation-specific arguments in a data field corresponding to a dimension not produced by the transformation instructions in the transformation data;
dispatch, to each of a plurality of the execution units associated with operations for which transformed local spaces have been generated, invocation data including said transformation data.