Extensible data format conversion techniques
AssigneesAdvanced Micro Devices, Inc., Xilinx, Inc.
InventorsStuart David Simpson Biles, Eric Ford Dellinger, Michael Estlick, Brian William Thompto
Life of this patent
Estimating the term…
Abstract
A processing system includes a processor to access an input data array in an initial format from a memory and convert the input data array into an output data array in a target format based on an instruction specifying an element size in the input data array. The converting includes unpacking data of the input data array based on the element size to generate unpacked data and employing a look up table (LUT) and the unpacked data to convert to the output data array in the target format. The processor also includes a plurality of compute units to execute operations of an artificial intelligence (AI) model with the output data array.
Claims20
A method comprising:
- accessing, by a parallel processor, an input data array in an initial format from a memory to execute one or more operations associated with an artificial intelligence (AI) model;
- converting, at conversion circuitry of the parallel processor, the input data array into an output data array in a target format based on an instruction specifying an element size in the input data array, wherein the converting comprises:
- unpacking data of the input data array based on the element size to generate unpacked data;
- and
- converting the unpacked data to the output data array in the target format;
- and
- executing the one or more operations of the AI model with the output data array at the parallel processor.
A method comprising:
- accessing, by a parallel processor, an input data array in an initial format from a memory to execute one or more operations associated with an artificial intelligence (AI) model, wherein the input data array comprises a plurality of elements;
- converting, at conversion circuitry of the parallel processor, the input data array into an output data array in a target format based on an instruction specifying an element size in the input data array, wherein the converting comprises:
- creating a vector of indices based on the element size;
- and
- unpacking data of the input data array into the vector of indices to generate unpacked data which is used with a look up table (LUT) to convert to the output data array in the target format;
- and
- executing the one or more operations of the AI model with the output data array at the parallel processor.
A processing system comprising:
- a processor to access an input data array comprising a plurality of elements in an initial format from a memory and convert the input data array into an output data array in a target format based on an instruction specifying an element size in the input data array, wherein the converting comprises:
- creating a vector of indices based on the element size;
- unpacking data of the input data array into the vector of indices to generate unpacked data;
- and
- converting the unpacked data to the output data array in the target format;
- and
- a plurality of compute units to execute operations of an artificial intelligence (AI) model with the output data array.
Dependencies are inferred from the claim text.