Buffered CPU-IP Data Interaction for High-Throughput Embedded Chips
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The frequent interactions between a central processing unit and IP modules in embedded chip systems limit the performance of the IP modules due to storage capacity constraints in register-based interactions.
Innovation Solution
Implementing a data interaction method that uses downlink and uplink buffers to store parameters and output data, respectively, reducing interactions by allowing the IP modules to acquire parameters from downlink buffers and store output data in uplink buffers, triggering bulk data reading when thresholds are reached, and managing buffer states to optimize parameter and data transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If register-based interactions are used between central processing unit and IP modules, then parameter configuration and state acquisition can be performed, but the storage capacity constraints limit the performance of IP modules due to frequent interactions
Solution Approach 1:
The patent introduces downlink buffers and uplink buffers as intermediary storage structures between the central processing unit and IP modules. The downlink buffer stores parameters to be written to IP modules, while the uplink buffer stores output data from IP modules. These buffers act as mediators that decouple the direct interaction between the CPU and IP modules, allowing batch operations and reducing the frequency of interactions while maintaining the ability to configure parameters and acquire states.
2Ease of operation
If frequent parameter configuration and state acquisition are performed, then the central processing unit can control IP modules, but the interaction frequency increases and burdens the system
Solution Approach 1:
The patent implements preliminary action by pre-storing multiple parameters in the downlink buffer before IP module execution. Instead of configuring parameters one-by-one during IP module operation, the central processing unit can prepare and store multiple parameters in advance in the downlink buffer. Similarly, the uplink buffer pre-allocates space for output data, allowing batch retrieval. This preliminary preparation reduces the number of real-time interactions needed during IP module execution, thereby reducing interaction time while maintaining full control capability.
3Reliability
If direct interaction between central processing unit and IP modules is maintained, then real-time control is achieved, but the performance of IP modules is limited due to storage capacity constraints
Solution Approach 1:
The patent segments the interaction interface into two distinct buffered channels: downlink for parameter input and uplink for output data. This segmentation allows independent management of parameter writing and data reading operations. The downlink buffer and uplink buffer can be filled and emptied independently, enabling the IP module to process data continuously while the central processing unit operates at its own pace. This segmentation maintains real-time control reliability while significantly improving data processing capability by eliminating interaction bottlenecks.
Data Source
AI summary
A data interaction method, apparatus and system, an electronic device, and a storage medium are disclosed. The method includes: obtaining, in response to a parameter writing operation signal, a first parameter and storing the first parameter in a downlink buffer; acquiring a second parameter from the downlink buffer and transmitting the second parameter to a data processing circuit unit, for the data processing circuit unit to process data based on the second parameter so as to obtain output data; storing the output data in an uplink buffer; recording the quantity of output data stored in the uplink buffer, and transmitting, when the quantity of the output data stored in the uplink buffer reaches an output quantity threshold, storage state information of the uplink buffer; and outputting the output data stored in the uplink buffer in bulk in response to a data reading operation instruction.