Method for avoiding overflow of multiplication of roots-of-unity by using overflow safe approximated values table
AssigneeLSI Corporation
InventorsAssaf Prihed, Shai Kalfon, Eran Goldstein
Life of this patent
Estimating the term…
Abstract
An apparatus including a processor, a computer readable storage medium, and a lookup memory. The computer readable storage medium generally contains computer executable instruction that when executed by the processor perform operations involving fixed point multiplication. The lookup memory generally stores values used in the fixed point multiplication. The values stored in the lookup memory are approximated based upon a predetermined value to prevent overflow in the fixed point multiplication.
Claims20
An apparatus comprising:
- a processor;
- a computer readable storage medium containing computer executable instruction that when executed by the processor perform operations involving fixed point multiplication;
- and
- a lookup memory storing fixed point representations of a plurality of roots-of-unity values, wherein said values are reduced from a true magnitude based upon a predetermined value to prevent overflow in the fixed point multiplication.
A method of preventing overflow in an operation involving multiplication of fixed-point representations of roots of unity values, comprising the steps of:
- calculating a number of floating point values for a number of roots of unity;
- reducing the floating point values based upon a predetermined value;
- translating the reduced floating point values to fixed point representations;
- and
- storing a lookup table embodying the fixed point representations of the floating point reduced values in a memory device.
An apparatus comprising:
- a lookup memory embodying fixed point representations of a plurality of roots-of-unity values, wherein one or both of an imaginary part and a real part of said roots-of-unity values are reduced from a respective true magnitude based upon a predetermined value to prevent overflow in a fixed point multiplication;
- and
- means for processing digital signals, wherein said digital signal processing means is configured to perform fixed point multiplication using said roots-of-unity values in said lookup memory.
Dependencies are inferred from the claim text.