Anchored Data Conversion for Reproducible Parallel Summation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Floating-point arithmetic calculations are non-associative, leading to reproducibility issues and difficulties in parallelizing code, especially in high-performance computing, where sums are not reproducible unless computed in the exact same order, causing problems in high-performance computing environments.
Innovation Solution
The introduction of a high-precision anchored (HPA) number format, which consists of a pair of a long two's complement integer and an anchor integer, allowing for associative and reproducible calculations by maintaining the anchor value unchanged during additions, enabling parallel processing without the need for sequential operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If floating-point arithmetic is used, then a large range of numeric values can be represented, but calculations become non-associative and sums are not reproducible
Solution Approach 1:
The patent segments floating-point numbers into two separate components: an integer part and a fractional part. The integer part is stored in an integer register and the fractional part is stored in a separate fractional register. This segmentation allows integer arithmetic (which is associative) to be applied to the integer part while preserving the fractional precision, thereby achieving both wide range representation and reproducible sums.
2Adaptability or versatility
If floating-point arithmetic is used, then numeric values over a large range can be represented, but parallel processing becomes difficult
Solution Approach 1:
By segmenting floating-point operations into integer and fractional components stored in separate registers, the patent enables parallel processing of integer parts using standard integer arithmetic units while fractional parts are handled separately. This allows multiple operations to proceed in parallel without the sequential constraints imposed by floating-point normalization and rounding.
3Reliability
If high-precision anchored numbers are used, then associative and reproducible calculations are achieved, but conversion from floating-point is required
Solution Approach 1:
The patent segments the floating-point conversion process into extracting the integer part and fractional part separately. The integer part is converted using standard integer conversion while the fractional part is handled through a dedicated fractional register. This segmentation simplifies the overall conversion circuitry compared to implementing full high-precision anchored number conversion, as each component can be converted independently using simpler logic.
Data Source
AI summary
An apparatus is provided, that includes an instruction decoder responsive to an anchored-data processing instruction, to generate one or more control signals. Conversion circuitry is responsive to the one or more control signals to perform a conversion from a data value to an anchored-data select value. The conversion is based on anchor metadata indicative of a given range of significance for the anchored-data select value. Output circuitry is responsive to the one or more control signals, to write the anchored-data select value to a register.