# US 2008/0082893 A1 · Error-tolerant multi-threaded memory systems with reduced error accumulation

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

- Number: US 2008/0082893 A1
- Document ID: US20080082893A1
- Status: Published application
- Kind: A1
- Application number: 11/863,353
- Filed: 2007-09-28
- Published: 2008-04-03
- Earliest claimed: 2006-09-29
- Assignees: University of Connecticut
- Inventors: Lei Wang
- Web page: https://patentgrep.com/patent/US20080082893A1

## CPC

- G06F 11/1008: {in individual solid state devices (G06F11/1004 takes precedence)}
- G06F 11/00: Error detection; Error correction; Monitoring (error detection, correction or monitoring in information storage based on relative movement between record carrier and transducer G11B20/18; monitoring, i.e. supervising the progress of recording or reproducing G11B27/36; in static stores G11C29/00)
- G06F 11/1076: {Parity data used in redundant arrays of independent storages, e.g. in RAID systems}
- G06F 11/1658: {Data re-synchronization of a redundant component, or initial sync of replacement, additional or spare unit}
- G06F 11/1666: {where the redundant component is memory or memory area}
- G06F 11/20: using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods establishing and/or utilizing an error-tolerant multithreaded register file are provided. The systems and methods employ dynamic multithreading redundancy (DMR) for error correction. Non-overlapped register access patterns associated create hardware redundancy dynamically that is exploited for error control. Immediate write-back and self-recovery techniques are employed to further enhance the error correction functionalities of the disclosed systems and methods. Error control is improved for memory components and processing functions in multithreaded computing systems.

## Claims

### Claim 1

A system for multithreaded computing, comprising:
one or more memory components configured to meet memory requirements of one or more multithreaded microprocessors;
error-correction means responsive to the detection of a transient or soft error in a memory location of the one or more memory components, the transient or soft error being associated with one or more threads; and
routing means associated with a multithreaded computing architecture that is operably connected to the one or more memory components and the error-correction means.

### Claim 2 (depends on 1)

The system for multithreaded computing of claim 1, wherein the one or more memory components are selected from the group consisting of multithreaded register files, cache memory and RAM.

### Claim 3 (depends on 1)

The system for multithreaded computing of claim 1, wherein the routing means is associated with a multithreaded computing architecture that includes a plurality of hardware components suitable for the execution of concurrent threads.

### Claim 4 (depends on 1)

The system for multithreaded computing of claim 1, wherein the error-correction means comprises dynamic multithreading redundancy (DMR) functionality.

### Claim 5 (depends on 1)

The system for multithreaded computing of claim 1, wherein the error-correction means is adapted to:
detect a soft error in data associated with an active thread;
detect availability of a redundant copy of the soft error-containing data associated with the active thread;
retrieve the redundant copy locally; and
use the retrieved redundant copy in place of the soft error-containing data in at least one computing operation.

### Claim 6 (depends on 1)

The system for multithreaded computing of claim 1, wherein the transient or soft error is selected from the group consisting of a transient error, a random error, an error caused by particle strike(s), an error caused by leakage, a charge-loss error, an error caused by noise, or a combination thereof.

### Claim 7 (depends on 1)

The system for multithreaded computing of claim 1, wherein the one or more memory components are associated with a single microprocessor.

### Claim 8 (depends on 1)

The system for multithreaded computing of claim 1, wherein the one or more memory components are associated with a plurality of microprocessors.

### Claim 9

A method of error correction in a multithreaded computer system, comprising:
detecting a transient or soft error in a memory location associated with an active thread;
detecting the availability of a redundant copy of error-containing data contained in the memory location associated with the active thread;
retrieving the redundant copy locally;
using the redundant thread copy of the error-containing data in place of the error-containing data in a computing operation to be executed; thereby achieving dynamic error correction in real time.

### Claim 10 (depends on 9)

The method of claim 9, further comprising initiating an immediate write-back error-control technique.

### Claim 11 (depends on 9)

The method of claim 9, further comprising initiating a self-recovery error-control technique.

### Claim 12 (depends on 9)

The method of claim 9, wherein the detecting step is conducted with respect to a multithreaded memory associated with a single microprocessor.

### Claim 13 (depends on 9)

The method of claim 9, wherein the detecting step is conducted with respect to a multithreaded memory associated with a plurality of microprocessors.

### Claim 14

A multithreading processing system, comprising:
at least two multithreaded register files, wherein each of the at least two multithreaded register files includes memory cells accessible by different threads;
means for establishing a flag bit with respect to the memory cells of each of the at least two multithreaded register files, the flag bit being established if a memory cell is free for reuse;
means for determining the status of the flag bit for a second of the at least two multithreaded register files when receiving a request to write to a first of the least two multithreaded register files, said means for status determination acting to write to both the first and second of the at least two multithreaded register files if the flag bit is established for the second multithreaded register file, thereby establishing a redundant data entry on the first and second multithreaded register files.

### Claim 15 (depends on 14)

The multithreaded processing system of claim 14, further comprising means for performing parity error checking with respect to a register or memory location associated with an active thread.

### Claim 16 (depends on 15)

The multithreaded processing system of claim 15, further comprising means for accessing the redundant data entry when the parity error checking means detects an error with respect to a register or memory location associated with an active thread.

### Claim 17 (depends on 16)

The multithreaded processing system of claim 16, further comprising means for updating the multithreaded register file associated with the active thread based on the redundant data entry when the parity error checking means detects an error with respect to the register or memory location associated with the active thread.

### Claim 18 (depends on 14)

The multithreaded processing system of claim 14, wherein the means for establishing a flag bit is adapted to configure the flag bits during runtime according to a free register list generated by a register renaming utility.

### Claim 19 (depends on 14)

The multithreaded processing system of claim 14, further comprising:
means for immediate write-back and self-recovery with respect to the redundant data entry.

### Claim 20 (depends on 14)

The multithreaded processing system of claim 14, wherein the at least two multithreaded register files are associated with a single microprocessor.

### Claim 21 (depends on 14)

The multithreaded processing system of claim 14, wherein the at least two multithreaded register files are associated with a plurality of microprocessors.

### Claim 22 (depends on 14)

The multithreaded processing system of claim 14, wherein each of the at least two multithreaded register files are selected from the group consisting of dual-threaded register files, triple-threaded register files, quad-threaded register files, and n-way threaded register files.
