8-bit floating point comparison instructions

A method, comprising:

• decoding a vector comparison instruction, the vector comparison instruction comprising a plurality of fields including a first source operand field to indicate a first plurality of 8-bit floating point source data elements, a second source operand field to indicate a second plurality of 8-bit floating point source data elements, a destination operand field to indicate a destination to store a corresponding plurality of 8-bit floating-point result data elements, and an opcode field to indicate a type of vector compare operation to be performed and a type of the 8-bit floating point source and result data elements, the type of 8-bit floating point source and result data elements comprising either 8-bit floating-point data elements having one sign bit with 4-bit exponents and 3-bit mantissas or 8-bit floating-point data elements having one sign bit with 5-bit exponents and 2-bit mantissas;
• and
• executing the decode vector comparison instruction in accordance with the opcode by:
• comparing a plurality of pairs of corresponding 8-bit floating-point source data elements of the first plurality of 8-bit floating-point source data elements and the second plurality of 8-bit floating point source data elements;
• and
• selecting an 8-bit floating-point source data element from each pair of the plurality of pairs of 8-bit floating-point source data elements based on the type of 8-bit floating-point vector compare operation, each selected 8-bit floating-point data element comprising a corresponding one of the plurality of 8-bit floating-point result data elements.

A system, comprising:

• memory to store a vector comparison instruction;
• decode circuitry to decode the vector comparison instruction, the vector comparison instruction comprising a plurality of fields including a first source operand field to indicate a first plurality of 8-bit floating point source data elements, a second source operand field to indicate a second plurality of 8-bit floating point source data elements, a destination operand field to indicate a destination to store a corresponding plurality of 8-bit floating-point result data elements, and an opcode field to indicate a type of vector compare operation to be performed and a type of the 8-bit floating point source and result data elements, the type of 8-bit floating point source and result data elements comprising either 8-bit floating-point data elements having one sign bit with 4-bit exponents and 3-bit mantissas or 8-bit floating-point data elements having one sign bit with 5-bit exponents and 2-bit mantissas;
• and
• execution circuitry to execute the vector comparison instruction in accordance with the opcode, the execution circuitry to:
• compare a plurality of pairs of corresponding 8-bit floating-point source data elements of the first plurality of 8-bit floating-point source data elements and the second plurality of 8-bit floating point source data elements;
• and
• select an 8-bit floating-point source data element from each pair of the plurality of pairs of 8-bit floating-point source data elements based on the type of 8-bit floating-point vector compare operation, each selected 8-bit floating-point data element comprising a corresponding one of the plurality of 8-bit floating-point result data elements.