Parallel Push and Relabel Path Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining the most energy-efficient path between two endpoints in a network with multiple possible paths require significant serial calculations, leading to substantial computational overhead and sub-optimal results due to local optimization efforts.
Innovation Solution
The implementation of a parallel processing system using a Push and Relabel algorithm on a graphics processing unit (GPU) to efficiently determine the globally optimized path by partitioning nodes into blocks and executing parallel relabel and push operations, allowing for wider SIMD instructions and increased thread parallelism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional serial processing methods are used to determine the global optimal path, then calculation accuracy is maintained, but computational time and power consumption increase significantly
Solution Approach 1:
The patent divides the graph nodes into multiple blocks that can be processed in parallel. Each block is independently processed by separate processing threads, allowing simultaneous computation of shortest paths from multiple source nodes. This segmentation enables the system to maintain calculation accuracy while significantly reducing computational time by exploiting parallel processing capabilities of GPU architecture.
Solution Approach 2:
The patent transitions from traditional CPU serial processing to GPU parallel processing, effectively adding a dimensional aspect of parallelism to the computation. By mapping the shortest path problem onto a parallel processing architecture with multiple streaming multiprocessors, the system achieves exponential speedup while maintaining the same algorithmic correctness, resolving the contradiction between accuracy and computational time.
2Measurement precision
If conventional serial processing methods are used to determine the global optimal path, then calculation accuracy is maintained, but power consumption increases significantly
Solution Approach 1:
By segmenting the computation into independent blocks processed in parallel, the system distributes the computational load across multiple GPU cores rather than overloading a single CPU core. This parallel distribution reduces the total energy consumption while maintaining the same level of calculation accuracy, as each processing unit operates at lower power levels simultaneously.
Solution Approach 2:
The patent replaces the traditional CPU-based sequential processing mechanism with a GPU-based parallel processing mechanism. This substitution leverages the fundamentally different architectural approach of GPUs, which are designed for parallel computation, thereby reducing power consumption for the same computational task while maintaining accuracy.
3Productivity
If local optimization efforts are applied at each juncture, then computational overhead is reduced, but path optimality deteriorates
Solution Approach 1:
The patent segments the graph into multiple processable blocks and applies the Bellman-Ford algorithm independently to each block in parallel. This segmentation allows the system to achieve both computational efficiency through parallel processing and path optimality by ensuring that the relaxation operation is thoroughly applied within each block, thus avoiding the sub-optimality of local greedy approaches.
Solution Approach 2:
The patent performs preliminary initialization of distance values and parent pointers for all nodes before the parallel relaxation process begins. This preliminary action ensures that when the parallel relaxation operations proceed, they start from a consistent and correct state, enabling both efficient parallel computation and guaranteed optimal path finding without requiring sequential processing.
Data Source
AI summary
Embodiments are generally directed to global optimal path determination utilizing parallel processing. An embodiment of an apparatus includes a central processing unit (CPU); a graphical processing unit (GPU), the GPU being capable of a plurality of processing threads; and a memory to store data for a system under evaluation, the system under evaluation including a set of nodes having a first endpoint, a second endpoint, and multiple paths between the first endpoint and the second endpoint. The apparatus is to determine a most energy efficient path between the first endpoint and the second endpoint utilizing parallel processing of a push and relabel graph cut algorithm. Performance of the push and relabel algorithm includes a plurality of process iterations, each process iteration including performance of a relabel operation, a push operation in a first direction, and a push operation in a second direction.