Reconfigurable Semiconductor Arithmetic Unit With Separate Fixed Value Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional semiconductor devices with reconfigurable circuits require significant labor, time, and high development costs due to fixed circuit configurations, and their processing speed is low because CPU direct control of arithmetic units leads to inefficient dynamic reconfiguration.
Innovation Solution
A semiconductor device with a sequencer instead of a CPU for controlling circuit configuration, featuring a configuration memory and a separate fixed value memory, allowing independent storage and retrieval of configuration and fixed value data, enabling dynamic reconfiguration of arithmetic units without the need for storing fixed values in the configuration memory.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If configuration information and fixed value data are stored in the same memory (configuration memory), then the device structure is simpler, but the storage capacity required increases and chip size increases
Solution Approach 1:
The patent divides the storage system into two separate memories: configuration memory for storing configuration information and fixed value memory for storing fixed value data. This segmentation allows each memory to be optimized for its specific purpose and reduces the total storage capacity required compared to a unified memory structure that would need to accommodate both types of data.
Solution Approach 2:
The patent introduces a new dimension of organization by separating storage by data type (configuration vs. fixed value) rather than storing all data in a single space. This dimensional change in storage organization enables more efficient memory utilization and reduces overall storage requirements.
2Ease of operation
If CPU directly controls arithmetic units for dynamic reconfiguration, then the device is easier to operate, but the processing speed is slow
Solution Approach 1:
The patent introduces a sequencer as an intermediary component between the CPU and the arithmetic units. The sequencer automatically controls the arithmetic units based on pre-stored configuration information in the configuration memory, eliminating the need for direct CPU intervention in each reconfiguration operation. This intermediary mechanism enables faster, automatic reconfiguration while maintaining ease of operation through programmatic control.
Solution Approach 2:
The system performs preliminary action by pre-storing configuration information and fixed value data in respective memories before execution. The sequencer then automatically retrieves and applies these pre-prepared configurations without real-time CPU intervention, enabling fast dynamic reconfiguration at clock cycle speeds while maintaining operational simplicity.
3Device complexity
If fixed value data is stored in configuration memory, then no separate fixed value memory is needed, but the configuration memory requires larger storage capacity
Solution Approach 1:
The patent segments the storage function into two specialized memories: configuration memory for control information and fixed value memory for data values. This segmentation reduces the total chip area compared to a single large unified memory, as each specialized memory can be optimized for its specific data type and size requirements.
Solution Approach 2:
The fixed value memory serves as a universal storage resource that can be shared across multiple arithmetic units and configuration states. This multi-functional storage approach reduces the total storage capacity needed compared to having dedicated fixed value storage in each arithmetic unit or in the configuration memory.
Data Source
AI summary
A semiconductor device includes a configuration memory for storing configuration data, an arithmetic unit whose circuit configuration can be reconfigured in accordance with the configuration data, and a fixed value memory for storing fixed value data to be supplied to the arithmetic unit. Since the configuration data and fixed value data to be supplied to the arithmetic unit are stored in the different memories, no data area for storing the fixed value data need be set in the configuration memory. This makes it possible to supply a predetermined fixed value to the arithmetic unit by storing only information for reading out fixed value data from the fixed value memory.