SIMD Lattice Computation via Dummy Node Insertion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing SIMD function in CPUs and GPUs cannot be used to speed up computation processing in the lattice method due to the varying number of nodes at different levels, preventing parallel vector operations.
Innovation Solution
A computing device with a SIMD function adds dummy nodes to enable vector operations by deciding the number and positions of dummy nodes to match the number of nodes at the previous level, allowing calculations to be performed using the SIMD function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the lattice method is used for numerical computation, then computation accuracy is improved, but the varying number of nodes at different levels prevents the use of SIMD functions, reducing processing speed
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the correspondence relationships between node values at different lattice levels before actual computation. This preprocessing step creates a lookup table that enables rapid retrieval during SIMD operations, resolving the contradiction by preparing data structures in advance that accommodate variable node counts while maintaining high-speed parallel processing capability
Solution Approach 2:
The patent introduces an intermediary data structure (correspondence relationship table) that mediates between the variable node counts at different lattice levels and the fixed-width SIMD registers. This intermediary layer maps variable-sized node value sets to fixed-size register operations, enabling SIMD acceleration without compromising the accuracy of lattice method computations
2Productivity
If dummy nodes are added to enable vector operations, then processing speed is improved through SIMD function usage, but device complexity increases due to additional nodes and correspondence management
Solution Approach 1:
The patent applies copying by creating virtual dummy nodes that replicate the structure needed for SIMD operations without adding physical computational resources. These dummy nodes are software-level abstractions that copy the necessary data patterns to fill SIMD registers, enabling vector operations while avoiding the hardware complexity of physically adding extra lattice nodes
Solution Approach 2:
The patent changes parameters by dynamically adjusting the correspondence relationships between node values based on the actual number of nodes at each lattice level. Rather than fixing the node structure, the system modifies the mapping parameters to match the variable node counts, allowing SIMD operations to adapt to different lattice configurations without increasing structural complexity
Data Source
AI summary
A computing device includes: a deciding unit which, in computation of values of nodes on a lattice in a direction where a value of m representing a horizontal axis coordinate of the lattice increases, decides dummy nodes to be added to m=n−1, so as to enable values of nodes on m=n to be calculated by adding the dummy nodes to m=n−1 and executing a vector operation through the use of the SIMD function by using values of nodes on m=n−1 and values of the added dummy nodes; an adding unit adding the dummy nodes decided by the deciding unit to m=n−1; and a calculating unit calculating the values of the nodes present on m=n by executing the vector operation through the use of the SIMD function by using the values of the nodes on m=n−1 and the values of the dummy nodes added by the adding unit.