# US 2020/0174750 A1 · Variable precision floating-point multiplier

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

- Number: US 2020/0174750 A1
- Document ID: US20200174750A1
- Status: Published application
- Kind: A1
- Application number: 16/785,933
- Filed: 2020-02-10
- Published: 2020-06-04
- Earliest claimed: 2016-08-22
- Assignees: Altera Corporation
- Applicants: Altera Corporation
- Inventors: Martin Langhammer
- Web page: https://patentgrep.com/patent/US20200174750A1

## CPC

- G06F 7/4876: {Multiplying}
- G06F 7/485: Adding; Subtracting {(G06F7/4833, G06F7/4836 take precedence)}
- G06F 2207/3816: Accepting numbers of variable word length
- G06F 2207/382: Reconfigurable for different fixed word lengths
- G06F 2207/483: Indexing scheme relating to group G06F7/483
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Integrated circuits with specialized processing blocks are provided. The specialized processing blocks may include floating-point multiplier circuits that can be configured to support variable precision. A multiplier circuit may include a first carry-propagate adder (CPA), a second carry-propagate adder (CPA), and an associated rounding circuit. The first CPA may be wide enough to handle the required precision of the mantissa. In a bridged mode, the first CPA may borrow an additional bit from the second CPA while the rounding circuit will monitor the appropriate bits to select the proper multiplier output. A parallel prefix tree operable in a non-bridged mode or the bridged mode may be used to compute multiple multiplier outputs. The multiplier circuit may also include exponent and exception handling circuitry using various masks corresponding to the desired precision width.

## Claims

### Claim 1

An integrated circuit, comprising:
a compressor tree circuit configured to generate signals; and
adder circuitry configured to receive the signals from the compressor tree circuit and operable to output corresponding floating-point values having exponents with exponent widths varying among at least three different values.

### Claim 2 (depends on 1)

The integrated circuit of claim 1, wherein the exponent widths vary consecutively within a range from a first value to a second value.

### Claim 3 (depends on 1)

The integrated circuit of claim 1, wherein the at least three different values comprise 5, 8, and some number greater than 5 and less than 8.

### Claim 4 (depends on 1)

The integrated circuit of claim 1, wherein the at least three different values comprise 5, 9, and some number greater than 5 and less than 9.

### Claim 5 (depends on 1)

The integrated circuit of claim 1, wherein the adder circuitry comprises carry propagate adder circuitry.

### Claim 6 (depends on 1)

The integrated circuit of claim 1, wherein the compressor tree circuit is configured to generate the signals based on a first input signal having a first exponent and a second input signal having a second exponent that is different than the first exponent, further comprising:
exponent handling circuitry configured to combine the first exponent with the second exponent based on an adjustable bias value that varies as a function of the precision of the first and second exponents.

### Claim 7 (depends on 6)

The integrated circuit of claim 6, wherein the exponent handling circuitry is further configured to combine the first exponent with the second exponent based on an adjustable normalization value.

### Claim 8 (depends on 1)

The integrated circuit of claim 1, further comprising:
exponent exception handling circuitry that includes a masking circuit that varies depending on the value of the exponents.

### Claim 9 (depends on 8)

The integrated circuit of claim 8, wherein the masking circuit comprises a first logic gate configured to detect whether the floating-point values are zero.

### Claim 10 (depends on 9)

The integrated circuit of claim 9, wherein the masking circuit comprises a second logic gate configured to detect whether the floating-point values are infinity or not-a-number (NaN).

### Claim 11

An integrated circuit, comprising:
a compressor tree circuit configured to generate signals; and
adder circuitry configured to receive the signals from the compressor tree circuit and operable to output corresponding floating-point values having a first precision in a first mode and having a second precision that is different than the first precision in a second mode.

### Claim 12 (depends on 11)

The integrated circuit of claim 11, further comprising:
a multiplexing circuit that is coupled to the adder circuitry and that is configured to select a first bit in the first mode and to select a second bit in a different position than the first bit in the second mode.

### Claim 13 (depends on 11)

The integrated circuit of claim 11, further comprising:
a rounding circuit configured to perform rounding based on at least two different carry signals generated from the adder circuitry.

### Claim 14 (depends on 11)

The integrated circuit of claim 11, wherein the adder circuitry comprises:
a first carry propagate adder; and
a second carry propagate adder, wherein the adder circuitry has a first rounding position located between the first and second carry propagate adders in the first mode and has a second rounding position that is different than the first rounding position in the second mode.

### Claim 15 (depends on 11)

The integrated circuit of claim 11, wherein the adder circuitry is configured to generate a first sum, a second sum that is equal to the first sum plus one, and a third sum that is equal to the first sum plus two.

### Claim 16 (depends on 11)

The integrated circuit of claim 11, wherein the compressor tree circuit is configured to generate the signals based on a first input signal having a first exponent and a second input signal having a second exponent that is different than the first exponent, further comprising:
exponent handling circuitry configured to combine the first exponent with the second exponent based on an adjustable bias value that varies as a function of the precision of the first and second exponents.

### Claim 17 (depends on 16)

The integrated circuit of claim 16, wherein the exponent handling circuitry is further configured to combine the first exponent with the second exponent based on an adjustable normalization value.

### Claim 18 (depends on 11)

The integrated circuit of claim 11, further comprising:
exponent exception handling circuitry that includes a masking circuit that varies depending on the value of exponents of the floating-point values.

### Claim 19 (depends on 18)

The integrated circuit of claim 18, wherein the masking circuit comprises a first logic gate configured to detect whether the floating-point values are zero.

### Claim 20 (depends on 19)

The integrated circuit of claim 19, wherein the masking circuit comprises a second logic gate configured to detect whether the floating-point values are infinity or not-a-number (NaN).
