Memory Frequency Control via Feed-Forward Compression Statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SoC systems face challenges in efficiently managing memory, memory bus, and system interconnect frequency, leading to suboptimal power consumption and performance due to the tradeoff between conserving power and maintaining performance, as existing methods either overestimate or underestimate memory bandwidth requirements.
Innovation Solution
The system employs feed-forward compression statistics to adjust the frequency and voltage settings of memory devices, memory buses, and system interconnects by generating compressed data buffers and related statistics, allowing accurate bandwidth voting and reducing power consumption without compromising performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If memory bandwidth vote is based on worst-case (highest) required bandwidth, then performance is maintained, but power consumption increases
Solution Approach 1:
The system performs preliminary compression of data buffers before memory access operations. By pre-compressing the data and determining actual compression ratios, the system establishes accurate bandwidth requirements in advance, avoiding the need to allocate worst-case bandwidth and thereby reducing power consumption while maintaining performance.
Solution Approach 2:
The system implements feedback mechanisms where compression statistics from previous operations are used to adjust and refine bandwidth voting for subsequent operations. This feedback loop allows the system to learn from actual data characteristics and optimize bandwidth allocation, preventing both over-provisioning (high power) and under-provisioning (performance degradation).
2Use of energy by moving object
If memory bandwidth vote is based on estimated typical bandwidth, then power consumption is reduced, but performance may degrade due to inaccurate estimation
Solution Approach 1:
The system replaces traditional mechanical estimation methods with information-theoretic compression analysis. Instead of using rough statistical estimates or fixed rules, the system uses actual compression algorithms to determine precise data sizes and bandwidth requirements, eliminating the uncertainty inherent in estimation-based approaches.
Solution Approach 2:
The system dynamically changes bandwidth voting parameters based on actual compression ratios observed in the data. By monitoring compression statistics and adjusting bandwidth allocations in response to these parameter changes, the system achieves accurate power-performance optimization without relying on static or estimated values.
Data Source
AI summary
Systems, methods, and computer programs are disclosed for controlling memory frequency. One method comprises a first memory client generating a compressed data buffer and compression statistics related to the compressed data buffer. The compressed data buffer and the compression statistics are stored in a memory device. Based on the stored compression statistics, a frequency or voltage setting of the memory device is adjusted for enabling a second memory client to read the compressed data buffer.