Near Memory Processor Scheduling via Real-Time State Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inefficient processing time and idle times occur when multiple processes with varying operation sizes are executed by near memory processors, as determining the optimal processor for each process is complex, leading to suboptimal overall operation time and resource utilization.
Innovation Solution
A processor-implemented method that receives operation requests, determines priority information for near memory processors based on state information such as write, read, and operation time states, and allocates requests to processors to minimize idle times and maximize resource utilization by considering priority, channel, and operation size information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If operation requests are allocated to near memory processors without considering real-time state information, then allocation is simple and fast, but idle times increase and resource utilization decreases
Solution Approach 1:
The patent implements a feedback mechanism where the scheduler continuously monitors state information from near memory processors (including operation status, temperature, and power consumption) and uses this real-time data to dynamically adjust allocation decisions. This closed-loop feedback system resolves the contradiction by making allocation both intelligent (improving utilization) and adaptive to current system conditions.
Solution Approach 2:
The scheduler performs preliminary actions by pre-evaluating the state of near memory processors and maintaining an updated state table before allocation requests arrive. This allows the system to quickly make informed allocation decisions without complex real-time calculations during the allocation moment itself, thus improving resource utilization while keeping the allocation process efficient.
2Productivity
If operation requests are allocated without considering processor state, then allocation is quick, but overall operation time increases due to idle times
Solution Approach 1:
The patent applies dynamics by making the allocation strategy adaptive and flexible based on real-time processor states. The scheduler dynamically adjusts allocation decisions according to current conditions such as processor availability, temperature, and power consumption levels. This dynamic approach minimizes idle times by continuously matching requests with the most suitable processors, thereby improving operation speed without sacrificing efficiency.
3Productivity
If multiple processes are executed without priority-based allocation, then allocation is straightforward, but resource utilization and operation efficiency decrease
Solution Approach 1:
The patent implements local quality by assigning different priority levels and allocation strategies to different near memory processors based on their specific state characteristics. Each processor is evaluated individually based on its current operation status, temperature, and power consumption, allowing the scheduler to make optimized allocation decisions for each local context rather than applying a uniform allocation policy.
4Productivity
If near memory processors handle operations without temperature and power consideration, then allocation is simple, but energy efficiency and thermal management deteriorate
Solution Approach 1:
The patent applies parameter changes by incorporating temperature and power consumption as key parameters in the allocation decision-making process. The scheduler monitors these physical parameters in real-time and adjusts allocation based on their values, directing requests to processors with favorable thermal and power characteristics. This resolves the contradiction by optimizing processing throughput while maintaining energy efficiency through parameter-aware scheduling.
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
A method and apparatus with process scheduling is provided. The method includes receiving operation requests from a plurality of processes; determining priority information of a plurality of near memory processors based on predetermined state information of a plurality of memories which correspond to the plurality of near memory processors; allocating the received operation requests to at least one near memory processor based on the determined priority information; and updating state information of at least one memory of the plurality of memories corresponding to the at least one near memory processor in a state table.