# US 8,280,936 B2 · Packed restricted floating point representation and logic for conversion to single precision float

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 8,280,936 B2
- Document ID: US8280936B2
- Status: Granted
- Kind: B2
- Application number: 11/648,265
- Filed: 2006-12-29
- Published: 2010-10-07
- Granted: 2012-10-02
- Earliest claimed: 2006-12-29
- Assignees: Intel Corporation
- Inventors: Hong Jiang
- Web page: https://patentgrep.com/patent/US8280936B2

## CPC

- G06F 7/38: Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F 9/30025: {Format conversion instructions, e.g. Floating-Point to Integer, decimal conversion}
- G06F 9/30036: {Instructions to perform operations on packed data, e.g. vector, tile or matrix operations}
- H03M 7/24: Conversion to or from floating-point codes
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H03M: CODING; DECODING; CODE CONVERSION IN GENERAL (using fluidic means F15C4/00; optical analogue/digital converters G02F7/00; coding, decoding or code conversion, specially adapted for particular applications, see the relevant subclasses, e.g. G01D, G01R, G06F, G06T, G09G, G10L, G11B, G11C, H04B, H04L, H04M, H04N; ciphering or deciphering for cryptography or other purposes involving the need for secrecy G09C)

## Abstract

An apparatus for expanding an immediate vector of restricted data structures may include logic connected to a first memory and a second memory connected to the logic. The first memory may store the immediate vector of restricted data structures that specify distinct floating point numbers. The immediate vector may have a fixed number of bits. The logic may expand the vector of restricted data structures into a number of corresponding expanded data structures that also specify the distinct floating point numbers. Each of the expanded data structures may also have the fixed number of bits. The second memory may store the number of corresponding expanded data structures.

## Claims

### Claim 1

An apparatus comprising:
a memory to store eight bit floating point data structures, each of said eight bit floating point data structures including one sign bit, three exponent bits, and four fraction bits; and
logic to populate exponent bits of an expanded data structure with multiple instances of an inversion of an exponent bit not having a value of zero in one of the eight bit floating point data structures unless a group of remaining bits of the one of eight bit floating point data structures all have a value of zero, and to populate the exponent bits of the expanded data structure with multiple instances of an inversion of an exponent bit having a value of zero in one of the eight bit floating point data structures unless the group of remaining bits of the one of eight bit floating point data structures all have the value of zero, the logic including:
an inverter with an input of one of the three exponent bits of the eight bit floating point data structure;
an OR gate with inputs of a remaining two of the three exponent bits and the four fraction bits of the eight bit floating point data structure; and
an AND gate with inputs of an output of the inverter and an output of the OR gate.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein, in each of the eight bit floating point data structures, the sign bit occupies a most significant bit, and the four fraction bits occupy four least significant bits.

### Claim 3 (depends on 1)

The apparatus of claim 1, wherein each of the eight bit floating point data structures specifies one of 256 different floating point numbers ranging from about −31 to 31.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein the expanded data structure consists of:
one sign bit,
eight exponent bits, and twenty-three fraction bits.

### Claim 5 (depends on 1)

The apparatus of claim 1, wherein the memory to store eight bit floating point data structures comprises a thirty-two bit memory buffer storing four different eight bit floating point data structures.

### Claim 6 (depends on 5)

The apparatus of claim 5, further comprising:
a 128 bit memory buffer storing four different expanded data structures, the four different expanded data structures resulting from expanding the four different eight bit floating point data structures.

### Claim 7

A method comprising:
storing an eight bit floating point data structure in a memory, said eight bit floating point data structure including one sign bit, three exponent bits, and four fraction bits;
populating, using a digital logic circuit, exponent bits of an expanded data structure with multiple instances of an inversion of an exponent bit not having a value of zero in the eight bit floating point data structure unless a group of remaining bits of the eight bit floating point data structure all have a value of zero, comprising:
providing a most significant bit of the three exponent bits of the data structure to an input of an inverter of the digital logic circuit;
providing two least significant bits of the three exponent bits and the four fraction bits of the data structure to inputs of an OR gate of the digital logic circuit; and
providing an output of the inverter and an output of the OR gate to inputs of an AND gate of the digital logic circuit.

### Claim 8 (depends on 7)

The method of claim 7 further comprising:
providing an output of the AND gate to second, third, fourth, fifth, and sixth most significant bits of eight exponent bits of the expanded data structure.

### Claim 9 (depends on 8)

The method of claim 8 further comprising:
transferring the most significant bit of the three exponent bits of the data structure to a most significant bit of the eight exponent bits of the expanded data structure; and
transferring the two least significant bits of the three exponent bits of the data structure to two least significant bits of the eight exponent bits of the expanded data structure.
