# US 2016/0253235 A1 · Recycling Error Bits in Floating Point Units

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

- Number: US 2016/0253235 A1
- Document ID: US20160253235A1
- Status: Published application
- Kind: A1
- Application number: 15/149,988
- Filed: 2016-05-09
- Published: 2016-09-01
- Earliest claimed: 2012-11-14
- Assignees: Intel Corporation
- Applicants: Intel Corporation
- Inventors: Shih-Lien L. Lu; Helia Naeimi; Ralph Nathan; Daniel Sorin
- Web page: https://patentgrep.com/patent/US20160253235A1

## CPC

- G06F 11/10: Adding special bits or symbols to the coded information, e.g. parity check, casting out 9's or 11's
- G06F 9/3001: {Arithmetic instructions}
- G06F 9/30014: {with variable precision}
- G06F 9/30101: {Special purpose registers}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A mechanism for recycling error bits in a floating point unit is disclosed. A system of the disclosure includes a memory and a processing device communicably coupled to the memory. In one embodiment, the processing device comprising a floating point unit (FPU) to generate a result value from applying an operation on floating point number inputs to the FPU and generate an error value using the result value. The FPU also writes the result value to a first register of the processing device dedicated to storing results from the operation of the FPU and writes the error value to a second register of the processing device dedicated to storing errors from the operation of the FPU. The second register is separate from the first register.

## Claims

### Claim 1

A system, comprising:
a memory; and
a processing device communicably coupled to the memory, the processing device comprising a floating point unit (FPU) to:
generate a result value from applying an operation on floating point number inputs to the FPU;
generate an error value using the result value;
write the result value to a first register of the processing device dedicated to storing results from the operation of the FPU; and
write the error value to a second register of the processing device dedicated to storing errors from the operation of the FPU, wherein the second register is separate from the first register.

### Claim 2 (depends on 1)

The system of claim 1, wherein the error value is accessible from the second register.

### Claim 3 (depends on 1)

The system of claim 1, wherein an instruction set architecture (ISA) of the processing device comprises an instruction to move the error value from the second register.

### Claim 4 (depends on 3)

The system of claim 3, wherein the instruction to move the error value from the second register is utilized by one or more software applications executed by the processing device to access the error value.

### Claim 5 (depends on 1)

The system of claim 1, wherein the second register is dedicated to one of a plurality of precisions of the FPU.

### Claim 6 (depends on 5)

The system of claim 5, wherein the plurality of precisions comprise at least one of a 32-bit precision, a 64-bit precision, and a 128-bit precision.

### Claim 7 (depends on 1)

The system of claim 1, wherein in response to the FPU supporting a packed N-bit architecture, the second register comprises k N/k bit floating point values, where k is specified in an instruction to the FPU to perform the operation.

### Claim 8 (depends on 1)

The system of claim 1, wherein the error value is a result of at least one of a rounding error of the operation or a binary/decimal conversion of the operation.

### Claim 9

A method, comprising:
generating, by a floating point unit (FPU) of a processing device, a result value from applying an operation on floating point number inputs to the FPU;
generating, by the FPU, an error value using the result value;
writing, by the FPU, the result value to a first register of the processing device dedicated to storing results from the operation of the FPU; and
writing, by the FPU, the error value to a second register of the processing device dedicated to storing errors from the operation of the FPU, wherein the second register is separate from the first register.

### Claim 10 (depends on 9)

The method of claim 9, wherein the error value is accessible from the second register.

### Claim 11 (depends on 9)

The method of claim 9, wherein an instruction set architecture (ISA) of the processing device comprises an instruction to move the error value from the second register.

### Claim 12 (depends on 11)

The method of claim 11, wherein the instruction to move the error value from the second register is utilized by one or more software applications executed by the processing device to access the error value.

### Claim 13 (depends on 9)

The method of claim 9, wherein the second register is dedicated to one of a plurality of precisions of the FPU.

### Claim 14 (depends on 13)

The method of claim 13, wherein the plurality of precisions comprise at least one of a 32-bit precision, a 64-bit precision, and a 128-bit precision.

### Claim 15 (depends on 9)

The method of claim 9, wherein in response to the FPU supporting a packed N-bit architecture, the second register comprises k N/k bit floating point values, where k is specified in an instruction to the FPU to perform the operation.

### Claim 16

A non-transitory machine-readable storage medium including data that, when accessed by a machine, cause the machine to perform operations comprising:
accessing, by a processing device of the machine, a first dedicated register for storing an error value resulting from a result value from applying a first floating point operation performed on input operands to a floating point unit (FPU) of the processing device, the FPU performing the first floating point operation on the input operands, wherein the result value is stored in a second dedicated register and the second dedicated register is separate from the first dedicated register; moving, by the processing device, the error value from the first dedicated register to a floating point register of the processing device; and
utilizing, by the processing device, the error value stored in the floating point register as part of a software instruction executed by the processing device.

### Claim 17 (depends on 16)

The non-transitory machine -readable storage medium of claim 16, wherein the error value is accessible from the first dedicated register.

### Claim 18 (depends on 16)

The non-transitory machine -readable storage medium of claim 16, wherein the error value is moved prior to executing a second floating operation written in the floating point register.

### Claim 19 (depends on 16)

The non-transitory machine -readable storage medium of claim 16, wherein the first dedicated register is associated with one of a plurality of precisions of the FPU.

### Claim 20 (depends on 16)

The non-transitory machine -readable storage medium of claim 16, wherein in response to the FPU supporting a packed N-bit architecture, the second register comprises k N/k bit floating point values, where k is specified in an instruction to the FPU to perform the operation.
