Software Selectable SIMD Parallelism Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SIMD processors consume excessive power due to continuous operation of all parallel elements, even when not all resources are needed, leading to inefficient power usage and performance issues in applications with varying degrees of parallelism.
Innovation Solution
Implementing a method to selectively activate and deactivate parallel processing elements based on software instructions, matching the degree of parallelism to the task requirements, by using control logic to power down unused elements and adjust data processing width dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If all parallel processing elements operate continuously to maintain maximum processing capability, then processing speed and performance are improved, but power consumption increases excessively
Solution Approach 1:
The patent implements dynamic adjustment of the number of active parallel processing elements based on the degree of parallelism required by the current data processing task. The system transitions from a static configuration where all elements operate continuously to a dynamic configuration where only the necessary number of elements are activated, allowing the processing capability to adapt to varying workload requirements and thereby reducing power consumption while maintaining performance when needed.
Solution Approach 2:
The patent changes the operational parameter of parallel processing elements from a fixed state to a variable state. By adjusting the number of active elements as a parameter based on task requirements, the system optimizes the balance between processing speed and power consumption. This parameter change enables the system to operate at full capacity when high performance is needed and reduce capacity when power conservation is prioritized.
2Productivity
If the data path width is increased to process larger data units, then processing efficiency for parallel data is improved, but power consumption increases due to more processing elements being required
Solution Approach 1:
The system dynamically adjusts the effective data path width by controlling the number of active parallel processing elements. Instead of maintaining a fixed wide data path that requires all elements to be powered, the system adapts the active width to match the actual processing needs, thereby reducing power consumption while preserving the ability to process large data units when necessary.
Solution Approach 2:
The patent segments the parallel processing elements into active and inactive groups based on the required degree of parallelism. This segmentation allows the system to activate only the necessary number of elements for the current task, effectively reducing the active data path width and associated power consumption while maintaining the capability to process larger data units when needed by activating additional segments.
3Adaptability or versatility
If all parallel processing elements remain active to handle maximum data parallelism, then processing capability is maintained, but power consumption increases unnecessarily for applications with low parallelism requirements
Solution Approach 1:
The patent implements dynamic adaptation of processing capability to match application requirements. The system monitors or determines the degree of parallelism needed for the current task and adjusts the number of active processing elements accordingly. This dynamic adaptation ensures that processing capability is maintained only when necessary, eliminating unnecessary power consumption in applications with low parallelism requirements while preserving full capability when needed.
Solution Approach 2:
The system changes the operational parameter of processing capability from a fixed maximum to a variable level. By adjusting the number of active elements as a parameter based on application needs, the system achieves adaptability that matches actual requirements. This parameter change allows the system to maintain full processing capability when high parallelism is needed while reducing capability and power consumption for applications with lower requirements.
Data Source
AI summary
Selective power control of one or more processing elements matches a degree of parallelism to requirements of a task performed in a highly parallel programmable data processor. For example, when program operations require less than the full width of the data path, a software instruction of the program sets a mode of operation requiring a subset of the parallel processing capacity. At least one parallel processing element, that is not needed, can be shut down to conserve power. At a later time, when the added capacity is needed, execution of another software instruction sets the mode of operation to that of the wider data path, typically the full width, and the mode change reactivates the previously shut-down processing element.