Apparatus and method to compute reciprocal approximations
AssigneeIntel Corporation
InventorsVinodh Gopal, Gilbert M. Wolrich, Wajdi K. Feghali
Life of this patent
Estimating the term…
Abstract
A method and apparatus for reducing memory required to store reciprocal approximations as specified in Institute of Electrical and Electronic Engineers (IEEE) standards such as IEEE 754 is presented. Monotonic properties of the reciprocal function are used to bound groups of values. Efficient bit-vectors are used to represent information in groups resulting in a very compact table representation about four times smaller than storing all of the reciprocal approximations in a table.
Claims20
An apparatus comprising:
- memory to store a plurality of encoded groups of reciprocal approximations;
- and
- logic coupled to the memory to select one of the encoded groups based on a first portion of an index and using the selected encoded group and a second portion of the index to compute a reciprocal approximation corresponding to the index;
- wherein the encoded group includes a minimum difference between successive reciprocal approximations in the encoded group, a bit mask vector and a base reciprocal approximation.
A method comprising:
- storing a plurality of encoded groups of reciprocal approximations in computer readable memory;
- selecting one of the encoded groups, using a processor coupled to said computer readable memory, based on a first portion of an index;
- and
- computing, using said processor, a reciprocal approximation corresponding to the index using the selected encoded group and a second portion of the index;
- wherein the encoded group includes a minimum difference between successive reciprocal approximations in the encoded group, a bit mask vector and a base reciprocal approximation.
An article including a computer readable memory having associated information, wherein the information, when accessed, results in a processor performing:
- storing a plurality of encoded groups of reciprocal approximations in computer readable storage memory;
- selecting one of the encoded groups based on a first portion of an index and using the selected encoded group and a second portion of the index to compute a reciprocal approximation corresponding to the index;
- wherein the encoded group includes a minimum difference between successive reciprocal approximations in the encoded group, a bit mask vector and a base reciprocal approximation.
A system comprising:
- a dynamic random access memory;
- and
- a processor coupled to the dynamic random access memory, the processor including an arithmetic logic unit, the arithmetic logic unit comprising:
- memory to store a plurality of encoded groups of reciprocal approximations;
- and
- logic coupled to the memory to select one of the encoded groups based on a first portion of an index and using the selected encoded group and a second portion of the index to compute a reciprocal approximation corresponding to the index;
- wherein the encoded group includes a minimum difference between successive reciprocal approximations in the encoded group, a bit mask vector and a base reciprocal approximation.
Dependencies are inferred from the claim text.