Floating-Point Operation Circuit with Shared Multiplier
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional processors optimize floating-point operations separately, leading to suboptimal performance and complex circuitry, as they do not effectively integrate multiplication, accumulation, and fused multiplication and accumulation operations simultaneously.
Innovation Solution
A floating-point number operation circuit and method that integrates multiplication, mac, and fused mac operations using a shared multiplication and addition circuit, with a selection and control circuit to manage rounding and operation type, optimizing these operations for improved processor performance and simpler circuit design.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional processors optimize floating-point operations separately, then each operation can be implemented with dedicated circuitry, but the overall processor performance is suboptimal and the circuit becomes complex
Solution Approach 1:
The patent combines multiplication, mac, and fused mac operations into a single integrated circuit. The multiplication unit is shared across all three operation types, and the addition unit handles both mac and fused mac operations. This merging eliminates the need for separate dedicated circuits for each operation, reducing overall circuit complexity while maintaining the ability to perform all operations efficiently
Solution Approach 2:
The circuit implements multi-functionality by designing a universal multiplication unit that can serve multiplication, mac, and fused mac operations. The addition unit is designed to handle both mac and fused mac operations. This universal design allows a single circuit to perform multiple floating-point operations, improving processor performance without proportionally increasing circuit complexity
2Measurement precision
If two rounding operations are conducted in each mac operation, then the mac operation can be performed, but the fused mac operation loses accuracy compared to mac operation
Solution Approach 1:
The circuit dynamically adapts its rounding behavior based on the operation type being performed. For mac operations, it applies rounding at both multiplication and addition stages. For fused mac operations, it performs a single rounding operation at the addition stage. This dynamic adjustment allows the circuit to optimize accuracy for mac operations while maintaining efficiency for fused mac operations, resolving the trade-off between precision and productivity
Data Source
AI summary
This invention discloses a floating-point number operation circuit and a method thereof. The floating-point number operation circuit is configured to perform a fused multiplication and accumulation (fused mac) operation or a multiplication and accumulation (mac) operation on a first operand, a second operand, and a third operand, or perform a multiplication operation on the first operand and the second operand. The floating-point number operation circuit includes two rounding circuits, a multiplication circuit, a selection circuit, a control circuit, and an addition circuit. The control circuit controls the scheduling of various operations and the use of resources on each calculation path.