Floating-Point Arithmetic Circuitry with Configurable Exponent Range
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Integrated circuits face challenges in efficiently handling floating-point numbers and results of arithmetic operations that fall outside the range of normalized precision, leading to loss of accuracy due to the need for additional and costly logic circuitry to support denormalized numbers, especially in multipliers where no circuitry can be reused or functionally combined.
Innovation Solution
The introduction of an extended exponent range with additional bits and a configurable bias in floating-point arithmetic circuitry, allowing for dynamic adjustment of the exponent range to prevent loss of precision during overflow or underflow, and enabling efficient handling of normalized and denormalized numbers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional logic circuitry is added to support denormalized numbers, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent applies dynamics by making the exponent range configurable and adjustable. The system can dynamically switch between a first exponent range for normalized numbers and a second, extended exponent range for denormalized numbers. This allows the circuitry to adapt its behavior based on the input data, providing high precision when needed while maintaining simpler operation for normal cases, thus resolving the contradiction between precision and complexity.
Solution Approach 2:
The patent changes the parameter of the exponent range from fixed to variable. By introducing a configurable exponent range that can be adjusted based on the floating-point numbers being processed, the system can extend precision for denormalized numbers without permanently increasing the complexity of the logic circuitry. The parameter change allows the same hardware to serve multiple precision requirements.
2Device complexity
If fixed-point operators are used in specialized processing blocks, then device complexity is reduced, but adaptability worsens
Solution Approach 1:
The patent implements universality by designing the specialized processing block to handle both fixed-point and floating-point operations using the same hardware resources. The configurable exponent range allows the block to adapt to different numerical formats, making the circuitry universal rather than dedicated to a single operation type. This resolves the contradiction by enabling floating-point capability without requiring separate dedicated hardware.
Solution Approach 2:
The system uses dynamic configuration of the exponent range to adapt between fixed-point and floating-point modes. The same processing block can switch its behavior based on the input data characteristics, providing flexibility and versatility while maintaining a relatively simple base circuitry design. This dynamic adaptability eliminates the need for completely separate fixed-point and floating-point hardware.
3Measurement precision
If extended exponent range is implemented, then measurement precision is improved, but manufacturing cost increases
Solution Approach 1:
The patent makes the extended exponent range functionality universal by implementing it as a configurable feature that can be enabled or disabled based on the specific application requirements. This allows the same hardware design to serve both high-precision and standard applications, reducing manufacturing complexity and cost for typical use cases while providing enhanced precision when needed. The multi-functionality avoids the need for completely separate high-precision hardware paths.
Solution Approach 2:
By making the exponent range a configurable parameter rather than a fixed hardware constraint, the system can adjust precision levels based on actual needs. This parameter-based approach allows manufacturers to produce a single standardized device that can operate in different precision modes, reducing manufacturing complexity and cost compared to producing separate devices for different precision requirements.
Data Source
AI summary
The present embodiments relate to integrated circuits with floating-point arithmetic circuitry that handles normalized and denormalized floating-point numbers. The floating-point arithmetic circuitry may include a normalization circuit and a rounding circuit, and the floating-point arithmetic circuitry may generate a first result in form of a normalized, unrounded floating-point number and a second result in form of a normalized, rounded floating-point number. If desired, the floating-point arithmetic circuitry may be implemented in specialized processing blocks.