test(assemblers): add DOF-based mass assembly tests

This commit is contained in:
Jukka Aho
2026-05-09 18:37:22 +03:00
parent 8815157cb1
commit 9dd24ce2ce
+381
View File
@@ -0,0 +1,381 @@
# This file is a part of JuliaFEM.
# License is MIT: see https://github.com/JuliaFEM/JuliaFEM.jl/blob/master/LICENSE.md
"""
Mass-matrix microkernel tests through the DOF-based assembler.
What the test file proves:
1. **Default kernel produces zero `M`.** A `ContinuumKernel` /
`HeatKernel` constructed without `density` / `heat_capacity` returns
`evaluate_mass_entry == 0` so both `apply_M!` and `assemble_M!`
produce a structural-zero `M`. Existing static-only tests stay valid.
2. **Consistent mass matrix is correct.** With unit density on a unit
cube, the row-sum of `M` (which equals `sum(M*1) = ∫ ρ dV = ρ·V`)
matches `density * volume` to round-off — for elasticity it sums
each component independently, for heat it sums the scalar field.
3. **`M` is symmetric, SPD on the active DOFs, and `apply_M!` matches
`M * x`** to round-off for several random `x`.
4. **Density scaling is linear.** Doubling `density` doubles every
entry of `M` exactly.
5. **Block-diagonal in components.** The elasticity mass matrix has
no `(α, β)` cross terms — `M[i_x, j_y] == 0` for any node pair.
6. **Zero allocations.** Both `apply_M!` and `assemble_M!` allocate
0 bytes per call after warmup, same hard contract as `apply_K!` /
`assemble!`.
7. **Cache reuse.** Calling `assemble!` then `assemble_M!` on the
same cache produces an independent `K` and `M`, both correct.
The combination of (2) and (4) verifies the `evaluate_mass_entry`
microkernel is *the* place to extend mass behaviour — variable density
materials drop in by overriding `evaluate_mass_entry` for a new kernel
type without touching the assembler.
"""
using Test
using JuliaFEM
using JuliaFEM: ContinuumFormulation, FullThreeD, Temperature, Vertex
using JuliaFEM: @DOFSet, DOF
using JuliaFEM: LinearElastic, Displacement, ContinuumKernel
using JuliaFEM: HeatConductivity, HeatKernel
using JuliaFEM: DOFBasedCOOAssembler, DOFBasedCOOCache
using JuliaFEM: apply_K!, apply_M!, assemble_M!, extract_system
using JuliaFEM: create_elements!
using LinearAlgebra
using SparseArrays
using Tensors
using Random
# ----------------------------------------------------------------------------
# Mesh helpers (mirror the other DOF-based test files; kept local so this
# file stays independent and the WARNINGS about helper redefinitions are
# expected when the suites run together).
# ----------------------------------------------------------------------------
function _build_hex8_box(nx::Int, ny::Int, nz::Int)
nodes = Vec{3,Float64}[]
nidx(i, j, k) = (i - 1) + (j - 1) * (nx + 1) + (k - 1) * (nx + 1) * (ny + 1) + 1
for k in 1:(nz + 1), j in 1:(ny + 1), i in 1:(nx + 1)
push!(nodes, Vec{3}((Float64(i - 1) / nx,
Float64(j - 1) / ny,
Float64(k - 1) / nz)))
end
conns = NTuple{8,UInt32}[]
for k in 1:nz, j in 1:ny, i in 1:nx
n1 = nidx(i, j, k)
n2 = nidx(i + 1, j, k)
n3 = nidx(i + 1, j + 1, k)
n4 = nidx(i, j + 1, k)
n5 = nidx(i, j, k + 1)
n6 = nidx(i + 1, j, k + 1)
n7 = nidx(i + 1, j + 1, k + 1)
n8 = nidx(i, j + 1, k + 1)
push!(conns, (UInt32(n1), UInt32(n2), UInt32(n3), UInt32(n4),
UInt32(n5), UInt32(n6), UInt32(n7), UInt32(n8)))
end
return Mesh{8,Hexahedron{8}}(nodes, conns)
end
"Set up DOF-based elasticity assembly with optional density."
function _setup_elasticity(mesh; density::Float64 = 0.0)
material = LinearElastic(E = 210e9, ν = 0.3)
kernel = ContinuumKernel(ContinuumFormulation{FullThreeD}(),
material, Displacement{3}();
density = density)
S = @DOFSet{u::DOF{Displacement{3}, Vertex}}
elements, dof_mgr = create_elements!(mesh, Element{Hexahedron{8}, Lagrange{1}, S})
asm = DOFBasedCOOAssembler()
cache = DOFBasedCOOCache(elements, dof_mgr, mesh, kernel)
return cache, asm, kernel, mesh
end
"Set up DOF-based heat assembly with optional heat capacity."
function _setup_heat(mesh; heat_capacity::Float64 = 0.0)
material = HeatConductivity(k = 50.2)
kernel = HeatKernel(ContinuumFormulation{FullThreeD}(),
material, Temperature();
heat_capacity = heat_capacity)
S = @DOFSet{T::DOF{Temperature, Vertex}}
elements, dof_mgr = create_elements!(mesh, Element{Hexahedron{8}, Lagrange{1}, S})
asm = DOFBasedCOOAssembler()
cache = DOFBasedCOOCache(elements, dof_mgr, mesh, kernel)
return cache, asm, kernel, mesh
end
# ----------------------------------------------------------------------------
# 1. Default kernels (no density) → structurally-zero M
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: default kernels return structural-zero M" begin
println("\n" * "=" ^ 70)
println("MASS MATRIX — defaults (no density / heat_capacity)")
println("=" ^ 70)
@testset "ContinuumKernel default density" begin
mesh = _build_hex8_box(2, 1, 1)
cache, asm, kernel, m = _setup_elasticity(mesh)
@test kernel.density == 0.0
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
@test maximum(abs, M) == 0.0
x = randn(cache.ndofs); y = zeros(cache.ndofs)
apply_M!(y, cache, asm, kernel, m, x)
@test all(iszero, y)
println(" ContinuumKernel density=0 M structural zero ✓")
end
@testset "HeatKernel default heat_capacity" begin
mesh = _build_hex8_box(2, 1, 1)
cache, asm, kernel, m = _setup_heat(mesh)
@test kernel.heat_capacity == 0.0
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
@test maximum(abs, M) == 0.0
x = randn(cache.ndofs); y = zeros(cache.ndofs)
apply_M!(y, cache, asm, kernel, m, x)
@test all(iszero, y)
println(" HeatKernel heat_capacity=0 M structural zero ✓")
end
end
# ----------------------------------------------------------------------------
# 2. Correctness: row-sum = ρ·V, symmetry, SPD, apply_M! matches M*x.
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: correctness (row-sum, symmetry, SPD, apply_M!)" begin
println("\n" * "=" ^ 70)
println("MASS MATRIX — CORRECTNESS (consistent M)")
println("=" ^ 70)
Random.seed!(20260508)
@testset "Heat: ρcp = 1, unit cube" begin
ρcp = 1.0
mesh = _build_hex8_box(2, 2, 2)
cache, asm, kernel, m = _setup_heat(mesh; heat_capacity = ρcp)
n = cache.ndofs
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
# Symmetric to round-off
@test maximum(abs, M - M') < 1e-12 * maximum(abs, M)
# Row-sum: sum(M*1) = ρcp · ∫ dV = ρcp · 1.0 (unit cube)
rowsum_total = sum(M * ones(n))
@test isapprox(rowsum_total, ρcp * 1.0; rtol = 1e-12)
# SPD: every diagonal positive, x' M x > 0 for several random x
@test all(>(0.0), diag(M))
for _ in 1:5
x = randn(n)
@test x' * M * x > 0.0
end
# apply_M! ≡ M * x to round-off
max_rel = 0.0
for _ in 1:5
x = randn(n)
y_ref = M * x
y_mf = zeros(n); apply_M!(y_mf, cache, asm, kernel, m, x)
rel = norm(y_mf - y_ref) / max(norm(y_ref), 1.0)
@test rel < 1e-12
max_rel = max(max_rel, rel)
end
println(" Heat ndof=$n rowsum=$(round(rowsum_total; sigdigits = 5)) " *
"(expected $(ρcp)) max(apply_M! vs M*x)=$(round(max_rel; sigdigits = 3))")
end
@testset "Elasticity: ρ = 1, unit cube" begin
ρ = 1.0
mesh = _build_hex8_box(2, 2, 2)
cache, asm, kernel, m = _setup_elasticity(mesh; density = ρ)
n = cache.ndofs
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
# Symmetric to round-off
@test maximum(abs, M - M') < 1e-12 * maximum(abs, M)
# Row-sum per component: each of the 3 displacement components
# independently has sum = ρ · V. Total row-sum is 3 · ρ · V.
rowsum_total = sum(M * ones(n))
@test isapprox(rowsum_total, 3 * ρ * 1.0; rtol = 1e-12)
@test all(>(0.0), diag(M))
max_rel = 0.0
for _ in 1:5
x = randn(n)
y_ref = M * x
y_mf = zeros(n); apply_M!(y_mf, cache, asm, kernel, m, x)
rel = norm(y_mf - y_ref) / max(norm(y_ref), 1.0)
@test rel < 1e-12
max_rel = max(max_rel, rel)
end
println(" Elast ndof=$n rowsum=$(round(rowsum_total; sigdigits = 5)) " *
"(expected $(3 * ρ)) max(apply_M! vs M*x)=$(round(max_rel; sigdigits = 3))")
end
end
# ----------------------------------------------------------------------------
# 3. Density scaling is linear.
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: density scaling is linear" begin
mesh = _build_hex8_box(2, 1, 1)
cache_a, asm_a, kernel_a, m_a = _setup_elasticity(mesh; density = 1.0)
cache_b, asm_b, kernel_b, m_b = _setup_elasticity(mesh; density = 2.5)
assemble_M!(cache_a, asm_a, kernel_a, m_a); M_a, _ = extract_system(cache_a)
assemble_M!(cache_b, asm_b, kernel_b, m_b); M_b, _ = extract_system(cache_b)
@test maximum(abs, M_b - 2.5 * M_a) < 1e-12 * maximum(abs, M_a)
end
# ----------------------------------------------------------------------------
# 4. Block-diagonal structure of the elasticity mass matrix.
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: elasticity mass is block-diagonal in components" begin
mesh = _build_hex8_box(1, 1, 1)
cache, asm, kernel, m = _setup_elasticity(mesh; density = 1.0)
n = cache.ndofs
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
Md = Matrix(M)
# DOF layout for displacement is (node, x), (node, y), (node, z) per node.
# `local_dof_layout(Element{Hex8, ...})` orders them this way, and
# `create_elements!` produces a global numbering matching that order.
# So DOF index `3*(node-1) + α` for component α ∈ {1,2,3}.
nnodes = length(m.nodes)
@test n == 3 * nnodes
# Cross-component blocks must be zero — pick a random pair of nodes
# and verify M[i_x, j_y], M[i_x, j_z], M[i_y, j_z] are all zero.
Random.seed!(20260508)
for _ in 1:5
i = rand(1:nnodes); j = rand(1:nnodes)
for (αi, αj) in ((1, 2), (1, 3), (2, 3), (2, 1), (3, 1), (3, 2))
row = 3 * (i - 1) + αi
col = 3 * (j - 1) + αj
@test Md[row, col] == 0.0
end
end
end
# ----------------------------------------------------------------------------
# 5. Zero-alloc + KA-untouched contract.
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: zero allocations (apply_M! + assemble_M!)" begin
println("\n" * "=" ^ 70)
println("MASS MATRIX — ZERO-ALLOC")
println("=" ^ 70)
@testset "Elast cube $(nx)×$(ny)×$(nz)" for (nx, ny, nz) in
[(1, 1, 1), (2, 1, 1), (3, 2, 2)]
mesh = _build_hex8_box(nx, ny, nz)
cache, asm, kernel, m = _setup_elasticity(mesh; density = 7850.0)
n = cache.ndofs
x = ones(n); y = zeros(n)
# warmup
assemble_M!(cache, asm, kernel, m)
apply_M!(y, cache, asm, kernel, m, x)
GC.gc()
a_asm = @allocated assemble_M!(cache, asm, kernel, m)
@test a_asm == 0
GC.gc()
a_mf = @allocated apply_M!(y, cache, asm, kernel, m, x)
@test a_mf == 0
nelems = length(m.connectivity)
println(" Elast $(nx)×$(ny)×$(nz) $(lpad(nelems,3)) elem " *
"$(lpad(n,4)) dof assemble_M!=$a_asm apply_M!=$a_mf")
end
@testset "Heat cube $(nx)×$(ny)×$(nz)" for (nx, ny, nz) in
[(1, 1, 1), (2, 1, 1), (3, 2, 2)]
mesh = _build_hex8_box(nx, ny, nz)
cache, asm, kernel, m = _setup_heat(mesh; heat_capacity = 3500.0)
n = cache.ndofs
x = ones(n); y = zeros(n)
assemble_M!(cache, asm, kernel, m)
apply_M!(y, cache, asm, kernel, m, x)
GC.gc()
a_asm = @allocated assemble_M!(cache, asm, kernel, m)
@test a_asm == 0
GC.gc()
a_mf = @allocated apply_M!(y, cache, asm, kernel, m, x)
@test a_mf == 0
nelems = length(m.connectivity)
println(" Heat $(nx)×$(ny)×$(nz) $(lpad(nelems,3)) elem " *
"$(lpad(n,4)) dof assemble_M!=$a_asm apply_M!=$a_mf")
end
end
# ----------------------------------------------------------------------------
# 6. Cache reuse: assemble! then assemble_M! produces correct K and M.
# ----------------------------------------------------------------------------
@testset "evaluate_mass_entry: cache reuse (K then M)" begin
println("\n" * "=" ^ 70)
println("MASS MATRIX — CACHE REUSE (assemble! then assemble_M!)")
println("=" ^ 70)
mesh = _build_hex8_box(2, 2, 2)
cache, asm, kernel, m = _setup_elasticity(mesh; density = 7850.0)
n = cache.ndofs
# 1. Assemble K, extract.
assemble!(cache, asm, kernel, m)
K, _ = extract_system(cache)
# 2. Then assemble M into the *same* cache, extract.
assemble_M!(cache, asm, kernel, m)
M, _ = extract_system(cache)
# K and M are independent SparseMatrixCSC instances now.
# K must be SPD on free DOFs; M must be SPD outright (positive
# diagonal, x' M x > 0).
@test maximum(abs, K - K') < 1e-9 * maximum(abs, K)
@test maximum(abs, M - M') < 1e-12 * maximum(abs, M)
# M's row-sum must still equal 3 ρ V (unaffected by the prior K assembly).
rowsum_total = sum(M * ones(n))
@test isapprox(rowsum_total, 3 * 7850.0 * 1.0; rtol = 1e-12)
# Quick sanity: K is *not* M (would imply the cache wasn't reset
# between the two assemblies).
@test maximum(abs, K - M) > 0.5 * maximum(abs, K)
println(" Hex8 2×2×2 ndof=$n K SPD ✓ M SPD ✓ rowsum(M)=" *
"$(round(rowsum_total; sigdigits = 5)) (expected $(3 * 7850.0))")
end