Module gpu_fem_assembly 

GPU-accelerated FEM matrix assembly (CPU mock implementation).

This module provides Finite Element Method (FEM) matrix assembly routines that mirror a GPU implementation. All operations run on the CPU via plain loops for portability.

The formulation assumes 2-node bar/rod elements in 1-D for simplicity, making the element stiffness matrix 2×2 and DOF management straightforward. The same patterns extend to 2-D and 3-D elements.

Structs

GpuFemMesh

A FEM mesh with element connectivity and material parameters.

Functions

gpu_all_element_stiffness

Return the element stiffness matrices for all elements (parallel mock).

gpu_apply_dirichlet

Apply Dirichlet boundary conditions by zeroing constrained DOF rows/cols.

gpu_assemble_global

Parallel element stiffness computation — returns all element matrices.

gpu_dot_product

Parallel reduction dot product: a · b.

gpu_element_stiffness

Compute the 2×2 element stiffness matrix for bar/rod element e.

gpu_residual

Compute the residual vector r = f − K·u in parallel.