GPU Hardware Redundant Multi-Threading for Reliable Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Redundant computations on central processing units (CPUs) impose significant performance overhead due to limited parallel data processing capabilities, necessitating an efficient method for data verification in electronic devices.
Innovation Solution
A system and method for verifying computation output using hardware-based processors, where instances of computation are generated and processed on hardware-based processors, with outputs verified against each other in a hardware-based environment to ensure accuracy, leveraging compute units and graphics processing units (GPUs) for parallel processing and fault tolerance without altering existing architecture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If redundant computations are performed on CPUs for data verification, then reliability is improved, but productivity deteriorates due to significant performance overhead
Solution Approach 1:
The patent creates redundant copies of computation instances on GPU hardware to verify data accuracy. Multiple identical computation instances are executed in parallel, and their outputs are compared to detect errors. This copying approach leverages GPU's parallel processing capability to perform verification without significant performance penalty, resolving the contradiction between reliability and productivity.
Solution Approach 2:
The patent replaces the mechanical sequential processing of CPUs with GPU's parallel processing architecture. By substituting CPU-based redundant computation with GPU-based parallel computation, the system achieves the same reliability goal with significantly reduced performance overhead, as GPUs are specifically designed for handling multiple computation instances simultaneously.
2Reliability
If spatial redundancy is used by duplicating hardware, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent makes existing GPU hardware multi-functional by using it for both primary computation and redundant verification. The same GPU compute units that perform normal processing are also utilized to execute redundant computation instances for verification. This universal usage of hardware resources achieves fault tolerance without adding separate dedicated verification hardware, thereby avoiding increased device complexity.
Data Source
AI summary
A system and method for verifying computation output using computer hardware are provided. Instances of computation are generated and processed on hardware-based processors. As instances of computation are processed, each instance of computation receives a load accessible to other instances of computation. Instances of output are generated by processing the instances of computation. The instances of output are verified against each other in a hardware based processor to ensure accuracy of the output.