APPARATUS AND METHOD FOR COHERENT ACCELERATED CONVERSION BETWEEN DATA REPRESENTATIONS

摘要

The present invention provides an apparatus and a method for converting tensor data. For example, according to an embodiment of the present invention, the method comprises: a step of fetching source tensor blocks of a source tensor data structure, wherein each source tensor block includes a plurality of source tensor data components having a first numeric representation, and the source tensor data structure includes a predefined structural arrangement of source tensor blocks; a step of converting the one or more source tensor blocks into one or more destination tensor blocks including a plurality of destination tensor data components having a second numeric representation different from the first numeric representation, wherein sets of one or more source tensor blocks are converted to one or more corresponding destination tensor blocks in a specified order based on the first and second numeric representations; and a step of storing each individual destination tensor block in a designated memory region to maintain coherency with the predefined structural arrangement of the source tensor blocks.