18123567. SELECTIVE RE-EXECUTION OF INSTRUCTION STREAMS FOR RELIABILITY simplified abstract (Google LLC)

SELECTIVE RE-EXECUTION OF INSTRUCTION STREAMS FOR RELIABILITY

Organization Name

Google LLC

Inventor(s)

Jonathan Charles Masters of Boston MA (US)

SELECTIVE RE-EXECUTION OF INSTRUCTION STREAMS FOR RELIABILITY - A simplified explanation of the abstract

This abstract first appeared for US patent application 18123567 titled 'SELECTIVE RE-EXECUTION OF INSTRUCTION STREAMS FOR RELIABILITY

The technology described in the patent application aims to detect silent data corruption by selectively re-executing instructions in a processor.

• The renamer in the processor identifies instructions that can be performed out of order.
• A reorder buffer tracks instructions to be re-executed for computing their value a second time and comparing it with the value from the first computation.
• Before retiring an instruction, the initial result of executing it the first time is compared with the re-execution result to detect any discrepancies that may indicate silent data corruption.

Potential Applications: - This technology can be applied in critical systems where data integrity is paramount, such as financial transactions, healthcare systems, and aerospace applications. - It can also be used in high-performance computing environments to ensure accurate and reliable processing of complex algorithms.

Problems Solved: - Detecting silent data corruption that may go unnoticed in traditional processing systems. - Ensuring the accuracy and reliability of data processing in critical applications.

Benefits: - Improved data integrity and reliability in processing systems. - Early detection of potential data corruption issues. - Enhanced overall system performance and efficiency.

Commercial Applications: Title: Advanced Data Integrity Processor for Critical Systems This technology can be commercialized for use in industries such as finance, healthcare, aerospace, and high-performance computing. It offers a competitive advantage by ensuring data integrity and reliability in critical applications.

Questions about the technology: 1. How does this technology compare to traditional error detection methods? - This technology goes beyond traditional error detection methods by selectively re-executing instructions to detect silent data corruption that may not be caught by standard error-checking techniques.

2. What are the potential implications of silent data corruption in critical systems? - Silent data corruption in critical systems can lead to incorrect results, financial losses, compromised patient safety, or system failures. This technology helps mitigate these risks by detecting and addressing data integrity issues proactively.

Original Abstract Submitted

The technology is generally directed to detecting silent data corruption by selectively re-executing instructions. A renamer may receive decoded instructions from an instruction cache. The renamer may identify one or more of the decoded instructions that can be performed out of order. The reorder buffer may be configured to track instructions that are to be re-executed for purposes of computing their value a second time and comparing the value from the second computation with a value from a first computation. Prior to retiring, or completing, an instruction, an initial result of executing the instruction a first time may be compared with a re-execution result of re-executing the instruction a second or nth time. If the comparison indicates there is a different initial result than the re-execution result, an indication of possible silent data corruption may be generated for the instruction.