Failsafe Computation via Scalar Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for ensuring failsafe computations in critical systems require multiple math libraries, hardware math accelerators, and complex error analysis, which increase complexity and overhead, and are prone to errors due to differences in data types and memory corruption.
Innovation Solution
A method and device for performing failsafe computations by executing a first calculation and a second calculation using a scalar, where the second calculation multiplies and divides the first result by the scalar, and comparing the results to detect errors, without the need for multiple libraries or accelerators, and with memory protection to prevent corruption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple math libraries or hardware co-processors are used to ensure failsafe computation, then computation error detection capability is improved, but device complexity increases
Solution Approach 1:
The patent merges the failsafe computation capability into a single math library by implementing scalar transformation techniques that allow one library to perform both the original computation and the scaled computation. This eliminates the need for multiple separate math libraries or hardware co-processors while maintaining error detection capability.
Solution Approach 2:
The patent makes a single math library universal by enabling it to perform multiple functions: the original computation and the scaled computation for error detection. The scalar transformation approach allows the same library to generate both results, reducing device complexity while maintaining reliability.
2Reliability
If different data types are used for redundant calculations, then computation error detection is improved, but manufacturing precision decreases due to different resolutions
Solution Approach 1:
The patent changes the parameter being transformed from data type to scalar value. Instead of using different data types (float, double) with different resolutions, the patent uses the same data type with different scalar multiplication factors. This maintains consistent precision while enabling error detection through scalar transformation and comparison.
3Reliability
If redundant calculations are coded into source code, then failsafe computation is achieved, but ease of operation decreases due to design and maintenance hazards
Solution Approach 1:
The patent implements self-service by having the compilation process automatically generate the redundant scaled calculations and comparison logic. The compiler or build system automatically creates the necessary code for scalar transformation and result comparison, eliminating the need for manual coding of redundant calculations and reducing maintenance burden.
4Reliability
If memory protection mechanisms are added to prevent corruption, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent uses a software-based copying approach where the scaled calculation results are stored in separate memory locations and compared logically. This avoids the need for complex hardware memory protection units while achieving similar reliability through software-based result verification and comparison.
Data Source
AI summary
A method and device for performing failsafe computation, and a method of compiling code to perform a failsafe computation are provided. The method includes performing a first calculation (212) to generate a first result (214). A second calculation (218) is performed using a scalar (216) and the first calculation (212) to generate a second result (220). The second calculation (218) includes multiplying the first calculation (212) by the scalar (216) to generate a scaled result, and dividing the scaled result by the scalar (216) to generate the second result (220). The first result (214) and the second result (220) are compared to determine if they are equivalent.