Method for setting a bit associated with each component of packed floating-pint operand that is normalized in SIMD operations
AssigneeIntel Corporation
InventorsSivakumar Makineni, Sunnhyuk Kimn, Gautam B. Doshi, Roger A. Golliver
Life of this patent
Estimating the term…
Abstract
A method is provided for loading a packed floating-point operand into a register file entry having one or more associated implicit bits. The packed floating point operand includes multiple component operands. Significand and exponent bits for each component operand are copied to corresponding fields of the register entry, and the exponent bits are tested to determine whether the component operand is normalized. An implicit bit corresponding to the component operand is set when the component operand is normalized.
Claims11
A method comprising:
- retrieving a packed floating-point operand for a register file entry;
- copying significand, exponent and sign bits from each operand of the packed floating point operands to the register file entry;
- testing exponent bits of each of the floating-point operands;
- and
- for each operand, setting a bit associated with the register file entry responsive to a state of the exponent bits of the operand.
A method for loading a packed operand into a floating-point register file entry, the method comprising:
- copying significand bits and exponent bits for each component operand of the packed operand to corresponding fields in the register file entry;
- determining if each component operand is normalized;
- and
- setting to a first logic state a bit associated with each component operand that is normalized.
A processor comprising:
- a data cache;
- a floating point register file including a plurality of entries, each entry having one or more associated bits to indicate if an operand stored in the entry is normalized;
- and
- a load converter to copy significand and exponent bits for each of a plurality of operands in a packed floating-point operand from the data cache to one of the register file entries, to test the exponent bits of each of operand of the packed floating point operand and to set a corresponding bit to a first or second logic state according to a result of the test.
Dependencies are inferred from the claim text.