Compare commits

..
Author SHA1 Message Date
Sohail Reddy 9a12ddc9c1 Updated apps to use new architecture 2026-02-25 15:19:10 -08:00
Sohail Reddy 1d57520c3c Added field tracking by ID and generalized interface 2026-01-27 10:24:09 -07:00
Sohail Reddy b7753cb4db fixed cmake 2026-01-06 18:59:44 -08:00
Sohail Reddy 842fec483f Merge branch 'master' into multiapp 2026-01-06 18:53:05 -08:00
Sohail Reddy 4b0eafd047 Fixed const correctness 2026-01-06 18:49:38 -08:00
Sohail Reddy e793c70058 Updated fsi app for constrained inner product on ess_dof 2025-11-19 14:28:38 -08:00
Sohail Reddy 1856912d63 Fixed style 2025-11-19 14:24:23 -08:00
Sohail Reddy 3b6817e36b Added custom inner product to fixed point relaxation methods 2025-11-19 14:24:04 -08:00
Sohail Reddy e4dcfb684f Merge branch 'ode-stage-solve' into multiapp 2025-11-18 23:26:55 -08:00
Sohail Reddy 7dcf77da54 Merge branch 'master' into multiapp 2025-11-18 23:25:23 -08:00
Sohail Reddy 3e823f1da7 updated morpher coupled bc 2025-11-18 23:16:06 -08:00
Sohail Reddy 9a2fee8b6f moved FPI relaxation member functions to cpp 2025-10-19 21:44:25 -07:00
Sohail Reddy 4b38bd79e9 clean up fsi example 2025-10-16 18:05:41 -07:00
Sohail Reddy d1378b8a7b bug fix fpisolver; reset relaxation each call 2025-10-16 18:04:17 -07:00
Sohail Reddy 3972b24f3b fixed ale velocity calculation 2025-10-15 22:19:59 -07:00
Sohail Reddy 48bd66640d added command line option for relaxation factor and fixed mpi_c_bool 2025-10-15 22:19:35 -07:00
Sohail Reddy a230b87a20 Fixed pure virtual function overide if GSLib not built 2025-10-14 18:31:08 -07:00
Sohail Reddy 5f9f782ad2 fixed style 2025-10-14 17:01:51 -07:00
Sohail Reddy 159a259f30 fixed redeclared variable in FPISolver 2025-10-14 16:56:20 -07:00
Sohail Reddy 67da324c26 Merge branch 'master' into multiapp 2025-10-14 16:28:08 -07:00
Sohail Reddy b5cf20e76c added cmake for multiapp miniapps 2025-10-14 16:26:00 -07:00
Sohail Reddy 1510492dfd fixed style 2025-10-14 16:24:23 -07:00
Sohail Reddy 360e662cf8 FSI miniapp and updated elasticity and navier-stokes apps 2025-10-14 16:22:53 -07:00
Sohail Reddy c1ec6a8c72 bug fix in navier miniapp for Step() function 2025-10-14 13:27:04 -07:00
Sohail Reddy 805499d799 bugfix for updating grid functions for transfer 2025-10-14 13:26:15 -07:00
Sohail Reddy 7b0330ffca Removed repetitive code in cht miniapp 2025-10-14 13:23:55 -07:00
Sohail Reddy 68fd37de1d Added completed conjugate heat transfer test case 2025-10-11 17:44:41 -07:00
Sohail Reddy 9f1d2ac249 reverted changes to navier miniapp 2025-10-11 17:43:24 -07:00
Sohail Reddy a594ab5a3a fixed namespace std use 2025-10-11 17:42:57 -07:00
Sohail Reddy 343a8b473e moved fixed point solver to solvers.cpp 2025-10-11 17:13:31 -07:00
Sohail Reddy 911709c619 fixed use of MFEM_USE_MPI in multiapp 2025-10-11 16:10:17 -07:00
Sohail Reddy c303956b9f fixed source field ownership and cleaning 2025-10-11 01:29:50 -07:00
Sohail Reddy c031ec4466 Fixed const correctness and added operator ownership in multiapp 2025-10-10 17:37:12 -07:00
Sohail Reddy 58c56ec9dc added LinkedFieldsCollection to manage linked fields for each application 2025-10-10 00:01:12 -07:00
Sohail Reddy 7db740ede3 Added documentation, updated class names for coupling schemes 2025-09-28 23:36:06 -07:00
Sohail Reddy 0436adfb1c updated fixed point solver 2025-09-28 23:31:33 -07:00
Sohail Reddy a6f2d8981a Merge branch 'master' into multiapp 2025-09-08 18:02:54 -07:00
Sohail Reddy 90edc198e3 update fsi app 2025-09-08 17:58:24 -07:00
Sohail Reddy c691c7c854 Formatted multiapp files and added documentation 2025-09-08 10:21:16 -07:00
Sohail Reddy 7c9105bd0e fixed style 2025-09-08 01:41:26 -07:00
Sohail Reddy 773bac0114 updated navier-stokes miniapp for ALE 2025-09-08 01:29:33 -07:00
Sohail Reddy 440db8189c Added option for ALE to VectorConvectionNLFIntegrator 2025-09-08 01:28:26 -07:00
Sohail Reddy 94efb4cd81 Updated multiapp interface for type-erased operator, field transfer and coupling scheme 2025-09-08 01:27:47 -07:00
Sohail Reddy ce3bdc0dfc Simplified mesh-morphing/diffusion app 2025-09-08 01:24:53 -07:00
Sohail Reddy aaf59d3a83 updated elasticity miniapp to use Schur complement 2025-09-08 01:22:39 -07:00
Sohail Reddy 3b3ffa7d7a updated native navier miniapp to use standard ODESolve::Step signature 2025-09-05 13:53:12 -07:00
Sohail Reddy 2898ec669c Updated native navier miniapp to own/not-own the boundary conditions 2025-09-04 17:19:45 -07:00
Sohail Reddy f34bfbbd8d Added field transfer based on FindPoints 2025-09-04 10:35:50 -07:00
Sohail Reddy 3cec2cba78 updated navier-stokes and elasticity examples 2025-09-03 16:34:45 -07:00
Sohail Reddy f769e2a7d7 updated morpher to diffuse displacement or velocity 2025-09-03 16:34:08 -07:00
Sohail Reddy 493674e650 updated navier-stokes miniapp to work with scaled pressure 2025-09-03 16:31:58 -07:00
Sohail Reddy 13632cde51 updated elasticity miniapp 2025-09-03 16:31:03 -07:00
Sohail Reddy ae63fda0fd Updated multiapp object for multi-stage coupling 2025-08-19 11:54:02 -07:00
Sohail Reddy 48b3e428fb clean up 2025-08-19 11:53:32 -07:00
Sohail Reddy 89dd191866 bug-fix: fixed point solver; updated relaxation schemes to clamp relaxation factors 2025-08-19 11:47:14 -07:00
Sohail Reddy 2a4f9184be Updated Navier-Stokes miniapp for multi-stage coupling and artificial compressibility 2025-08-19 11:40:47 -07:00
Sohail Reddy ff66dce92e updated elasticity and morpher miniapps for multi-stage coupling 2025-08-19 11:38:25 -07:00
Sohail Reddy 01ef886faf Added parallelized DistanceSquared for Vectors 2025-08-19 11:09:47 -07:00
Sohail Reddy 4be67907c4 Working FSI miniapp 2025-08-13 16:12:04 -07:00
Sohail Reddy 9443523dec Updated the multiapp (CoupledApplication) framework 2025-08-13 16:10:19 -07:00
Sohail Reddy fac6dfe283 Added mesh morphing app 2025-08-13 16:09:50 -07:00
Sohail Reddy 7776766ee7 updated FPISolver 2025-08-13 16:09:09 -07:00
Sohail Reddy 8e24224a18 updated cht example...unused atm 2025-08-13 16:08:33 -07:00
Sohail Reddy fa4c607622 updated Navier-Stokes miniapp for partial assembly 2025-08-13 16:07:09 -07:00
Sohail Reddy 29bd541a3e updated elasticity miniapp 2025-08-13 16:03:23 -07:00
Sohail Reddy 369f82adad Added function to fetch TimeDependantOperator from ODESolver 2025-08-13 16:01:39 -07:00
Sohail Reddy e2bc1aefd8 Working elasticity miniapp 2025-08-01 19:09:05 -07:00
Sohail Reddy 033814fe70 fixed code style 2025-07-31 10:14:48 -07:00
Sohail Reddy f4d4a5c4da Working fully implicit incompressible navier-stokes solver 2025-07-31 10:13:33 -07:00
Sohail Reddy 6b8d341478 Renamed some multiapp member functions 2025-07-31 09:56:27 -07:00
Sohail Reddy 7a02905d32 Updated constructors for some bilinear and nonlinearform integrator to now take integration rule as input 2025-07-31 09:53:16 -07:00
Sohail Reddy a4c2ad2f3c Updated type erasure to support non-mfem apps 2025-07-21 10:27:35 -07:00
Sohail Reddy 4a11684323 Working monolithic (linear and nonlinear) solves for multiapp 2025-07-17 18:22:41 -07:00
Sohail Reddy 7b4b2de092 Merge branch 'master' into multiapp 2025-07-15 16:37:35 -07:00
Sohail Reddy 285aa03049 conv-diff mulitapp example 2025-07-15 16:35:56 -07:00
Sohail Reddy 27af36adc5 Changed class naming and fixed point solver and relaxation schemes 2025-07-15 16:35:03 -07:00
Sohail Reddy de781c05b8 Fixed function pointer for relaxation method in FPISolver 2025-07-10 17:04:58 -07:00
Sohail Reddy 125f5c9e48 Added relaxation methods to fixed point solvers and updated multiapp for explicit coupling 2025-07-10 16:59:03 -07:00
Sohail Reddy 0d6c8fbbf1 Updated multiapp partition implicit solve to use fixed point Solver object 2025-07-09 19:48:04 -07:00
Sohail Reddy 34ebc0f158 Changed scheme naming convention and added/separated Addtive and Alternating Schwarz coupling schemes 2025-07-08 18:16:32 -07:00
Sohail Reddy 8db7813d86 Working paritioned implicit solves 2025-07-08 11:06:27 -07:00
Sohail Reddy d31e5c8af0 Merge branch 'master' into multiapp 2025-07-02 20:46:57 -07:00
Sohail Reddy d7e43cfc74 added test case for two-way partitioned coupled solve 2025-07-02 20:45:51 -07:00
Sohail Reddy 82193e0404 Merge branch 'master' into multiapp 2025-06-16 15:09:01 -07:00
Sohail Reddy 5796633307 updated example for implicit time-integration 2025-06-16 15:05:22 -07:00
Sohail Reddy 3d55c021e3 updated header copyright 2025-06-12 14:47:11 -07:00
Sohail Reddy a30303ecf9 fixed type checking with SFINAE 2025-06-12 14:31:05 -07:00
Sohail Reddy 9287488024 Merge branch 'master' into multiapp 2025-06-11 14:11:32 -07:00
Sohail Reddy 9aa2866b0d Merge branch 'master' into multiapp 2025-05-29 14:00:03 -07:00
Sohail Reddy ad16d2f5ae added field transfer, multiapp abstractions and working one-way coupling 2025-05-29 13:58:58 -07:00
Sohail Reddy c552925a54 Merge branch 'master' into multiapp 2025-03-13 15:41:44 -07:00
Sohail Reddy c3b9b62336 updated mfem_application to perform index controlled operators 2025-02-20 11:28:02 -08:00
Sohail Reddy c36e5facd1 initial multiapp skeleton with MFEMApplication class and type-erased multiapp 2024-10-02 17:11:14 -07:00
361 changed files with 13328 additions and 38930 deletions
@@ -94,16 +94,6 @@ inputs:
description: If true, do not set any CXXFLAGS or LDFLAGS.
default: false
# Unfortunately, "uses:" fields cannot have references to variables like
# ${{env.MFEM_ACTIONS_VERSION}}, so the branch/tag name has to be hard coded.
# Therefore, in the future, when updating the version of the
# mfem/github-actions to use, we'll have to replace:
# - all definitions of MFEM_ACTIONS_VERSION and
# - all "uses:" fields that refer to mfem/github-actions.
MFEM_ACTIONS_VERSION:
description: Version (branch or tag) of the mfem/github-actions to use.
default: v2.7
runs:
using: 'composite'
steps:
@@ -128,7 +118,6 @@ runs:
echo UBSAN_LDFLAGS=${{inputs.UBSAN_LDFLAGS}} >> $GITHUB_ENV
echo MSAN_CXXFLAGS=${{inputs.MSAN_CXXFLAGS}} >> $GITHUB_ENV
echo MSAN_LDFLAGS=${{inputs.MSAN_LDFLAGS}} >> $GITHUB_ENV
echo MFEM_ACTIONS_VERSION=${{inputs.MFEM_ACTIONS_VERSION}} >> $GITHUB_ENV
shell: bash
- name: Env (dir)
+3 -3
View File
@@ -25,7 +25,7 @@ runs:
steps:
- uses: ./.github/actions/sanitize/config
- uses: actions/cache@v5
- uses: actions/cache@v4
if: ${{env.DEBUG == 'true'}}
id: debug
with:
@@ -53,7 +53,7 @@ runs:
run: echo CXXFLAGS=${{env.CXXFLAGS}} ${{env.UBSAN_CXXFLAGS}} >> $GITHUB_ENV
shell: bash
- uses: mfem/github-actions/build-mfem@v2.7
- uses: mfem/github-actions/build-mfem@v2.5
if: ${{steps.debug.outputs.cache-hit != 'true'}}
env:
CXXFLAGS: ${{env.CXXFLAGS}}
@@ -82,7 +82,7 @@ runs:
run: find . -type f -name '*.o' -delete
shell: bash
- uses: actions/upload-artifact@v7
- uses: actions/upload-artifact@v4
with:
name: build-${{inputs.par}}-${{inputs.sanitizer}}
path: mfem/build
-6
View File
@@ -12,11 +12,6 @@
name: 'Install MPI'
description: 'Installs MPI and set up its environment variables'
inputs:
NO_FLAGS:
description: If true, do not set any CXXFLAGS or LDFLAGS.
default: false
runs:
using: 'composite'
steps:
@@ -32,7 +27,6 @@ runs:
shell: bash
- name: Env (bis)
if: ${{ inputs.NO_FLAGS != 'true' }}
run: |
echo CXXFLAGS=${{env.CXXFLAGS}} ${{env.MPI_INC}} >> $GITHUB_ENV
echo LDFLAGS=${{env.LDFLAGS}} ${{env.MPI_LIB}} >> $GITHUB_ENV
+2 -2
View File
@@ -36,7 +36,7 @@ runs:
steps:
- uses: ./.github/actions/sanitize/config
- uses: actions/cache@v5
- uses: actions/cache@v4
if: ${{env.DEBUG == 'true' && inputs.cache-skip != 'true'}}
id: debug
with:
@@ -49,7 +49,7 @@ runs:
par: ${{inputs.par}}
sanitizer: ${{inputs.sanitizer}}
- uses: actions/download-artifact@v8
- uses: actions/download-artifact@v4
with:
name: build-${{inputs.par}}-${{inputs.sanitizer}}
path: mfem/build
+7 -7
View File
@@ -23,7 +23,7 @@ inputs:
runs:
using: 'composite'
steps:
- uses: actions/cache/restore@v5 # Cache for LLVM libcxx
- uses: actions/cache/restore@v4 # Cache for LLVM libcxx
with:
path: ${{env.LLVM_DIR}}
fail-on-cache-miss: true
@@ -32,32 +32,32 @@ runs:
- uses: ./.github/actions/sanitize/mpi
if: ${{inputs.par == 'true'}}
- uses: actions/cache/restore@v5 # Cache for Hypre
- uses: actions/cache/restore@v4 # Cache for Hypre
if: ${{inputs.par == 'true'}}
with:
path: ${{env.HYPRE_DIR}}
fail-on-cache-miss: true
key: ${{runner.os}}-ompi-build-${{env.HYPRE_DIR}}-int32-fp64-${{env.MFEM_ACTIONS_VERSION}}
key: ${{runner.os}}-ompi-build-${{env.HYPRE_DIR}}-int32-fp64-v2.5
- uses: actions/cache/restore@v5 # Cache for Metis
- uses: actions/cache/restore@v4 # Cache for Metis
if: ${{inputs.par == 'true'}}
with:
path: ${{env.METIS_DIR}}
fail-on-cache-miss: true
key: ${{runner.os}}-build-${{env.METIS_DIR}}-${{env.MFEM_ACTIONS_VERSION}}
key: ${{runner.os}}-build-${{env.METIS_DIR}}-v2.5
- name: Hypre/Metis links
if: ${{inputs.par == 'true'}}
run: ln -s -f ${{env.HYPRE_DIR}} hypre && ln -s -f ${{env.METIS_DIR}} metis-4.0
shell: bash
- uses: actions/cache/restore@v5 # Cache for LSAN suppression file
- uses: actions/cache/restore@v4 # Cache for LSAN suppression file
with:
path: ${{env.LSAN_DIR}}
fail-on-cache-miss: true
key: build-lsan-suppression-file
- uses: actions/checkout@v6 # Checkout the repository
- uses: actions/checkout@v4 # Checkout the repository
with:
path: mfem
# ref: ${{env.BRANCH}}
-42
View File
@@ -1,42 +0,0 @@
# MFEM Pull Request Review Agent Guide
## Purpose and scope
Review MFEM PRs for correctness, maintainability, performance, portability, test coverage, and MFEM consistency. Use the diff and PR context; reference source files, tests, and CI results when available. Follow `CONTRIBUTING.md`, especially Developer Guidelines, PR rules, checklist, and testing.
## Critical review pillars
- Correctness and numerical behavior
- API and user-facing impact
- Performance implications
- Maintainability and portability
## Review workflow
1. Read the PR description, linked issues, and intended behavior.
2. Inspect the diff before commenting.
3. Identify affected MFEM components, examples, tests, build or docs changes, and downstream APIs.
4. Analyze the code against the critical review pillars.
5. Compare the change against nearby code and MFEM patterns; flag unmotivated deviations.
6. Check whether tests and documentation were updated appropriately.
7. Review CI results and suggest actions.
8. Produce a structured review with prioritized findings.
9. Always limit conclusions to available evidence.
## MFEM-specific review checklist
- Component-aware scope: identify the touched subsystem (FEM, solvers, preconditioners, linear algebra, mesh, examples, miniapps, build, or docs) and assess its impact against the review pillars.
- Numerical and algorithmic behavior: assess issues in convergence, stability, tolerances, precision, iteration limits, and failure handling. If clear opportunities exist to improve the algorithmic approach, call them out with expected impact.
- API and user-facing impact: assess backward compatibility, user-visible behavior and default changes, migration impact, deprecations, and whether documentation clearly explains user-facing API changes.
- Data structure and memory semantics: assess ownership, lifetime, aliasing, container behavior, and device-host synchronization.
- Parallel and serial behavior: assess whether the change preserves equivalent semantics in serial and parallel modes where applicable; if logic is currently mode-specific, check whether extension to the other mode is straightforward (clear abstractions, no hard-wired assumptions), document constraints, and call out expected behavior differences explicitly.
- Backend and portability impact: assess likely cross-backend risks in CPU, CUDA, HIP, OCCA, RAJA, partial assembly, fallback paths, compiler compatibility, and platform assumptions.
- Build, dependency, and configuration impact: assess CMake or make changes, optional dependency behavior, and feature-flag interactions.
- Tests and docs alignment: check available regression or unit coverage evidence for changed behavior, and ensure docs are updated for new flags, APIs, options, or behavior changes.
- MFEM developer-guideline fit: keep code lean, simple, general, logically separated, and portable; suggest C++17 improvements when they clearly improve safety, clarity, or maintainability.
- New source files, examples, or miniapps: if a PR adds source/header files, verify they are properly wired into the relevant `makefile` and `CMakeLists.txt`, referenced in docs where applicable (including `doc/CodeDocumentation.dox`), and added to top-level `.gitignore` only when generated artifacts require it.
- Changelog: verify `CHANGELOG` is updated if the PR introduces significant new features or user-facing changes.
- MFEM conventions: use `real_t`; use `mfem::out`/`mfem::err` instead of `std::cout`/`std::cerr` in library code; flag large/binary files; if AI assistance is apparent but undisclosed, suggest following `CONTRIBUTING.md`.
- Edge cases: if the PR touches complex or error-prone areas, suggest additional tests for edge cases, failure modes, and parallel behavior.
## Commenting guidelines
- Keep comments concise, actionable, and grounded in the diff.
- Focus on correctness, behavior changes, and user impact over style nits.
- Be professional, concise, collaborative, technically precise, and avoid unsupported assumptions.
+7 -3
View File
@@ -13,7 +13,7 @@ Note that some of these scripts use the shared MFEM GitHub Actions from the exte
<https://github.com/mfem/github-actions>
For a particular action, e.g. `mfem/github-actions/build-mfem@v2.5`, the `v2.5` suffix denotes the branch (or tag) in the above from which the action is taken.
For a particular action, e.g. `mfem/github-actions/build-mfem@v2.5`, the `v2.5` suffix denotes the branch in the above from which the action is taken.
The current CI workflows are:
@@ -29,12 +29,16 @@ Runs a number of static repository-level sanity checks.
- `branch-history` guards against accidental commits of large files using the `--history` option of the `config/githooks/pre-push` script.
## `mfem-analysis.yml` (`build-analysis`)
Checks if the code builds and satisfies minimal requirements.
- `gitignore` builds hypre, METIS, and MFEM using `mfem/github-actions/build-hypre`, `mfem/github-actions/build-metis`, and `mfem/github-actions/build-mfem` and checks for correct `.gitignore` settings by running the `tests/scripts/gitignore` script.
## `builds-and-tests.yml`
Runs a matrix of builds and tests runs with different compilers, OS, mfem/hypre settings, etc. Also processes and upload Codecov reports.
One matrix job runs `tests/scripts/gitignore` after `make test-noclean` to check generated artifacts against `.gitignore`.
Uses the following GitHub Actions from <https://github.com/mfem/github-actions>:
- `mfem/github-actions/build-hypre`
+1 -1
View File
@@ -43,7 +43,7 @@ jobs:
remove-docker-images: 'true'
- name: Checkout
uses: actions/checkout@v6
uses: actions/checkout@v4
# It's easier to reference named variables than indexes of the matrix
- name: Set Environment
+28 -31
View File
@@ -40,7 +40,6 @@ env:
METIS_ARCHIVE_MAC: metis-4.0.3-mac.tgz
METIS_TOP_DIR: metis-4.0.3
MFEM_TOP_DIR: mfem
MFEM_ACTIONS_VERSION: v2.7
# Note for future improvements:
#
@@ -111,7 +110,6 @@ jobs:
build-system: make
hypre-target: int64
precision: fp64
gitignore-check: YES
- os: ubuntu-latest
target: opt
codecov: NO
@@ -155,7 +153,7 @@ jobs:
# /home/runner/work/mfem/mfem/mfem
# Note: Done now to access "install-hypre" and "install-metis" actions.
- name: checkout mfem
uses: actions/checkout@v6
uses: actions/checkout@v4
with:
path: ${{ env.MFEM_TOP_DIR }}
# Fetch the complete history for codecov to access commits ID
@@ -172,6 +170,20 @@ jobs:
env
shell: bash
# For info on Xcode see:
# - https://github.com/actions/runner-images/issues/12541
# - https://github.com/actions/runner-images/blob/releases/macos-15-arm64/20250811/images/macos/macos-15-arm64-Readme.md#xcode
- name: Xcode version setup (MacOS)
if: matrix.os == 'macos-latest'
run: |
XCODE_PATH="/Applications/Xcode_16.4.app"
echo "> sudo xcode-select -s ${XCODE_PATH}"
sudo xcode-select -s ${XCODE_PATH}
echo "> g++ -v"
g++ -v
echo "> clang++ -v"
clang++ -v
# Only get MPI if defined for the job.
# TODO: It would be nice to have only one step, e.g. with a dedicated
# action, but I (@adrienbernede) don't see how at the moment.
@@ -213,14 +225,14 @@ jobs:
- name: cache hypre
id: hypre-cache
if: matrix.mpi == 'par'
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{ env.HYPRE_TOP_DIR }}
key: ${{ runner.os }}-ompi-build-${{ env.HYPRE_TOP_DIR }}-${{ matrix.hypre-target }}-${{ matrix.precision }}-${{ env.MFEM_ACTIONS_VERSION }}
key: ${{ runner.os }}-ompi-build-${{ env.HYPRE_TOP_DIR }}-${{ matrix.hypre-target }}-${{ matrix.precision }}-v2.5
- name: get hypre
if: matrix.mpi == 'par' && steps.hypre-cache.outputs.cache-hit != 'true' && matrix.os != 'windows-latest'
uses: mfem/github-actions/build-hypre@v2.7
uses: mfem/github-actions/build-hypre@v2.5
with:
archive: ${{ env.HYPRE_ARCHIVE }}
dir: ${{ env.HYPRE_TOP_DIR }}
@@ -230,7 +242,7 @@ jobs:
- name: get hypre (Windows)
if: matrix.mpi == 'par' && steps.hypre-cache.outputs.cache-hit != 'true' && matrix.os == 'windows-latest'
uses: mfem/github-actions/build-hypre@v2.7
uses: mfem/github-actions/build-hypre@v2.5
with:
archive: ${{ env.HYPRE_ARCHIVE }}
dir: ${{ env.HYPRE_TOP_DIR }}
@@ -243,14 +255,14 @@ jobs:
- name: cache metis
id: metis-cache
if: matrix.mpi == 'par' && matrix.os != 'windows-latest'
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{ env.METIS_TOP_DIR }}
key: ${{ runner.os }}-build-${{ env.METIS_TOP_DIR }}-${{ env.MFEM_ACTIONS_VERSION }}
key: ${{ runner.os }}-build-${{ env.METIS_TOP_DIR }}-v2.5
- name: install metis
if: matrix.mpi == 'par' && matrix.os != 'windows-latest' && steps.metis-cache.outputs.cache-hit != 'true'
uses: mfem/github-actions/build-metis@v2.7
uses: mfem/github-actions/build-metis@v2.5
with:
archive: ${{ matrix.os != 'macos-latest' && env.METIS_ARCHIVE || env.METIS_ARCHIVE_MAC }}
dir: ${{ env.METIS_TOP_DIR }}
@@ -258,7 +270,7 @@ jobs:
- name: cache vcpkg (Windows)
id: vcpkg-cache
if: matrix.os == 'windows-latest'
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: vcpkg_cache
key: ${{ runner.os }}-${{ matrix.mpi }}-vcpkg-v1
@@ -283,8 +295,7 @@ jobs:
export HOMEBREW_NO_INSTALL_CLEANUP=1
brew update
brew install enzyme
ENZYME_LLVM=$(brew info enzyme | sed -n 's/^Required.*:.*\(llvm[^ ]*\).*/\1/p')
echo "ENZYME_LLVM=$ENZYME_LLVM"
ENZYME_LLVM=$(brew info enzyme | sed -n 's/^Required:.*\(llvm[^ ]*\).*/\1/p')
LLVM_PREFIX=$(brew --prefix $ENZYME_LLVM)
echo "LLVM_PREFIX=$LLVM_PREFIX" >> $GITHUB_ENV
echo "OMPI_CC=$LLVM_PREFIX/bin/clang" >> $GITHUB_ENV
@@ -292,7 +303,7 @@ jobs:
# MFEM build and test
- name: build
uses: mfem/github-actions/build-mfem@v2.7
uses: mfem/github-actions/build-mfem@v2.5
env:
VCPKG_DEFAULT_BINARY_CACHE: ${{ github.workspace }}/vcpkg_cache
with:
@@ -318,13 +329,7 @@ jobs:
- name: tests
if: matrix.build-system == 'make' && (matrix.target == 'opt' || matrix.os == 'ubuntu-latest')
run: |
cd ${{ env.MFEM_TOP_DIR }}
if [[ "${{ matrix.gitignore-check }}" == "YES" ]]; then
make test-noclean
else
make test
fi
shell: bash
cd ${{ env.MFEM_TOP_DIR }} && make test
- name: cmake checks
if: matrix.build-system == 'cmake' && matrix.target == 'dbg'
@@ -369,16 +374,8 @@ jobs:
# Code coverage (process and upload reports)
- name: codecov
if: matrix.codecov == 'YES'
uses: mfem/github-actions/upload-coverage@v2.7
uses: mfem/github-actions/upload-coverage@v2.5
with:
name: ${{ matrix.os }}-${{ matrix.build-system }}-${{ matrix.target }}-${{ matrix.mpi }}-${{ matrix.hypre-target }}-${{ matrix.precision }}
name: ${{ matrix.os }}-${{ matrix.build-system }}-${{ matrix.target }}-${{ matrix.mpi }}-${{ matrix.hypre-target }}
project_dir: ${{ env.MFEM_TOP_DIR }}
directories: "fem general linalg mesh"
env:
CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
- name: gitignore
if: matrix.gitignore-check == 'YES'
run: |
cd ${{ env.MFEM_TOP_DIR }}/tests/scripts
./runtest gitignore
+4 -14
View File
@@ -14,19 +14,9 @@ name: "Static Analysis"
on:
push:
branches: ["master", "next"]
paths-ignore: &docs-only-paths
- "**/*.md"
- "doc/**"
- ".binder/**"
- "CITATION.cff"
- "LICENSE"
- "NOTICE"
- "CHANGELOG"
- "INSTALL"
pull_request:
# The branches below must be a subset of the branches above
branches: ["master"]
paths-ignore: *docs-only-paths
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
@@ -50,11 +40,11 @@ jobs:
steps:
- name: Checkout repository
uses: actions/checkout@v6
uses: actions/checkout@v4
# Initializes the CodeQL tools for scanning.
- name: Initialize CodeQL
uses: github/codeql-action/init@v4
uses: github/codeql-action/init@v2
with:
languages: ${{ matrix.language }}
# If you wish to specify custom queries, you can do so here or in a config file.
@@ -67,7 +57,7 @@ jobs:
# Autobuild attempts to build any compiled languages (C/C++, C#, or Java).
# If this step fails, then you should remove it and run the build manually (see below)
- name: Autobuild
uses: github/codeql-action/autobuild@v4
uses: github/codeql-action/autobuild@v2
# ️ Command-line programs to run using the OS shell.
# 📚 See https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idstepsrun
@@ -80,4 +70,4 @@ jobs:
# ./location_of_script_within_repo/buildscript.sh
- name: Perform CodeQL Analysis
uses: github/codeql-action/analyze@v4
uses: github/codeql-action/analyze@v2
+100
View File
@@ -0,0 +1,100 @@
# Copyright (c) 2010-2025, Lawrence Livermore National Security, LLC. Produced
# at the Lawrence Livermore National Laboratory. All Rights reserved. See files
# LICENSE and NOTICE for details. LLNL-CODE-806117.
#
# This file is part of the MFEM library. For more information and source code
# availability visit https://mfem.org.
#
# MFEM is free software; you can redistribute it and/or modify it under the
# terms of the BSD-3 license. We welcome feedback and contributions, see file
# CONTRIBUTING.md for details.
name: "Build Analysis"
permissions:
actions: write
on:
push:
branches:
- master
- next
pull_request:
workflow_dispatch:
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
env:
HYPRE_ARCHIVE: v2.19.0.tar.gz
HYPRE_TOP_DIR: hypre-2.19.0
METIS_ARCHIVE: metis-4.0.3.tar.gz
METIS_TOP_DIR: metis-4.0.3
COVERAGE_ENV: mfem-coverage
jobs:
gitignore:
runs-on: ubuntu-latest
steps:
- name: checkout MFEM
uses: actions/checkout@v4
with:
path: mfem
- name: Get MPI (Linux)
run: |
sudo apt-get install openmpi-bin libopenmpi-dev
export OMPI_MCA_rmaps_base_oversubscribe=1
- name: Cache Hypre Install
id: hypre-cache
uses: actions/cache@v4
with:
path: ${{ env.HYPRE_TOP_DIR }}
key: ${{ runner.os }}-ompi-build-${{ env.HYPRE_TOP_DIR }}-v2.5
- name: Get Hypre
if: steps.hypre-cache.outputs.cache-hit != 'true'
uses: mfem/github-actions/build-hypre@v2.5
with:
archive: ${{ env.HYPRE_ARCHIVE }}
dir: ${{ env.HYPRE_TOP_DIR }}
target: int32
- name: Cache Metis Install
id: metis-cache
uses: actions/cache@v4
with:
path: ${{ env.METIS_TOP_DIR }}
key: ${{ runner.os }}-build-${{ env.METIS_TOP_DIR }}-v2.5
- name: Install Metis
if: steps.metis-cache.outputs.cache-hit != 'true'
uses: mfem/github-actions/build-metis@v2.5
with:
archive: ${{ env.METIS_ARCHIVE }}
dir: ${{ env.METIS_TOP_DIR }}
# MFEM build and test
- name: build-mfem
uses: mfem/github-actions/build-mfem@v2.5
with:
os: ${{ runner.os }}
target: opt
codecov: NO
mpi: par
build-system: make
hypre-dir: ${{ env.HYPRE_TOP_DIR }}
metis-dir: ${{ env.METIS_TOP_DIR }}
mfem-dir: mfem
- name: test (no clean)
run: |
cd mfem && make test-noclean
- name: gitignore
run: |
cd mfem/tests/scripts
./runtest gitignore
+4 -4
View File
@@ -38,7 +38,7 @@ jobs:
github.event.pull_request.head.repo.full_name != github.repository)
steps:
- name: checkout mfem
uses: actions/checkout@v6
uses: actions/checkout@v4
- name: copyright check
id: copyright
@@ -93,7 +93,7 @@ jobs:
github.event.pull_request.head.repo.full_name != github.repository)
steps:
- name: checkout mfem
uses: actions/checkout@v6
uses: actions/checkout@v4
- name: get astyle
run: |
@@ -110,7 +110,7 @@ jobs:
github.event.pull_request.head.repo.full_name != github.repository)
steps:
- name: checkout mfem
uses: actions/checkout@v6
uses: actions/checkout@v4
- name: get doxygen and graphviz
run: |
@@ -135,7 +135,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: checkout mfem
uses: actions/checkout@v6
uses: actions/checkout@v4
with:
fetch-depth: 0
+4 -8
View File
@@ -17,24 +17,20 @@ jobs:
runs-on: ubuntu-latest
name: 2.19.0
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/config
with:
NO_FLAGS: true
- name: Cache
id: cache
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{env.HYPRE_DIR}}
key: ${{runner.os}}-ompi-build-${{env.HYPRE_DIR}}-int32-fp64-${{ env.MFEM_ACTIONS_VERSION }}
key: ${{runner.os}}-ompi-build-${{env.HYPRE_DIR}}-int32-fp64-v2.5
- name: Setup
if: steps.cache.outputs.cache-hit != 'true'
uses: ./.github/actions/sanitize/mpi
with:
NO_FLAGS: true
- name: Build
if: steps.cache.outputs.cache-hit != 'true'
uses: mfem/github-actions/build-hypre@v2.7
uses: mfem/github-actions/build-hypre@v2.5
with:
archive: ${{env.HYPRE_TGZ}}
dir: ${{env.HYPRE_DIR}}
+2 -2
View File
@@ -27,13 +27,13 @@ jobs:
llvm_use_sanitizer: "Undefined"
name: ${{matrix.sanitizer}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/config
with:
NO_FLAGS: true
- name: Cache
id: cache
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{env.LLVM_DIR}}
key: build-libcxx-${{env.LLVM_VER}}-${{matrix.sanitizer}}
+2 -2
View File
@@ -17,11 +17,11 @@ jobs:
runs-on: ubuntu-latest
name: lsan.supp
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/config
- name: Cache
id: cache
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{env.LSAN_DIR}}
key: build-lsan-suppression-file
+4 -8
View File
@@ -17,24 +17,20 @@ jobs:
runs-on: ubuntu-latest
name: 4.0.3
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/config
with:
NO_FLAGS: true
- name: Cache
id: cache
uses: actions/cache@v5
uses: actions/cache@v4
with:
path: ${{env.METIS_DIR}}
key: ${{runner.os}}-build-${{env.METIS_DIR}}-${{env.MFEM_ACTIONS_VERSION}}
key: ${{runner.os}}-build-${{env.METIS_DIR}}-v2.5
- name: Setup
if: steps.cache.outputs.cache-hit != 'true'
uses: ./.github/actions/sanitize/mpi
with:
NO_FLAGS: true
- name: Build
if: steps.cache.outputs.cache-hit != 'true'
uses: mfem/github-actions/build-metis@v2.7
uses: mfem/github-actions/build-metis@v2.5
with:
archive: ${{env.METIS_TGZ}}
dir: ${{env.METIS_DIR}}
+9 -9
View File
@@ -28,7 +28,7 @@ jobs:
build:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/mfem
with:
par: ${{inputs.par}}
@@ -40,7 +40,7 @@ jobs:
env:
ex: ${{inputs.par && 'ex1p' || 'ex1'}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
@@ -58,7 +58,7 @@ jobs:
env:
exclude: ${{inputs.par && '-E "_ser"' || ''}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
@@ -82,7 +82,7 @@ jobs:
env:
exclude: ${{inputs.par && '-E "_ser"' || ''}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
@@ -107,7 +107,7 @@ jobs:
run: ${{inputs.par && '-R "_cpu_np"' || ''}}
exclude: ${{inputs.par && '"unit_tests|debug"' || '"^unit_tests$|debug"'}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
@@ -131,7 +131,7 @@ jobs:
env:
unit_tests: ${{inputs.par && 'punit_tests' || 'unit_tests'}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
@@ -146,7 +146,7 @@ jobs:
if: ${{steps.restore.outputs.cache-hit != 'true'}}
working-directory: mfem/build/tests/unit
run: find . -type f -name '*.o' -delete
- uses: actions/upload-artifact@v7
- uses: actions/upload-artifact@v4
with:
name: tests-${{inputs.par}}-${{inputs.sanitizer}}
path: mfem/build/tests/unit/${{env.unit_tests}}
@@ -165,14 +165,14 @@ jobs:
unit_tests: ${{inputs.par && 'punit_tests' || 'unit_tests'}}
np: ${{inputs.par && '_np=2' || ''}}
steps:
- uses: actions/checkout@v6
- uses: actions/checkout@v4
- uses: ./.github/actions/sanitize/restore
id: restore
with:
par: ${{inputs.par}}
sanitizer: ${{inputs.sanitizer}}
cache-path: mfem/build/tests/unit/${{env.unit_tests}}
- uses: actions/download-artifact@v8
- uses: actions/download-artifact@v4
if: ${{steps.restore.outputs.cache-hit != 'true'}}
with:
name: tests-${{inputs.par}}-${{inputs.sanitizer}}
-10
View File
@@ -17,17 +17,7 @@ permissions:
on:
push:
branches: ["master", "next"]
paths-ignore: &docs-only-paths
- "**/*.md"
- "doc/**"
- ".binder/**"
- "CITATION.cff"
- "LICENSE"
- "NOTICE"
- "CHANGELOG"
- "INSTALL"
pull_request:
paths-ignore: *docs-only-paths
workflow_dispatch:
concurrency:
-14
View File
@@ -313,8 +313,6 @@ miniapps/nurbs/nurbs_ex1
miniapps/nurbs/nurbs_ex1p
miniapps/nurbs/nurbs_ex3
miniapps/nurbs/nurbs_ex5
miniapps/nurbs/nurbs_ex10
miniapps/nurbs/nurbs_ex10p
miniapps/nurbs/nurbs_ex11p
miniapps/nurbs/nurbs_ex24
miniapps/nurbs/nurbs_solenoidal
@@ -340,14 +338,7 @@ miniapps/nurbs/nurbs_naca_cmesh
miniapps/nurbs/naca-cmesh.mesh
miniapps/nurbs/glvis_naca-cmesh.mesh
miniapps/nurbs/Naca_cmesh
miniapps/nurbs/nurbs_mesh_info
miniapps/nurbs/k*_*.dat
miniapps/nurbs/*-Surface.mesh
miniapps/nurbs/*.mesh
miniapps/nurbs/*.sol
miniapps/nurbs/deformed.*
miniapps/nurbs/elastic_energy.*
miniapps/nurbs/velocity.*
miniapps/performance/ex1
miniapps/performance/ex1p
@@ -369,7 +360,6 @@ miniapps/shifted/lsf_integral
miniapps/tools/display-basis
miniapps/tools/load-dc
miniapps/tools/convert-dc
miniapps/tools/compare-dc
miniapps/tools/gridfunction-bounds
miniapps/tools/lor-transfer
miniapps/tools/plor-transfer
@@ -443,10 +433,6 @@ miniapps/diag-smoothers/mg-abs-l1-jacobi
miniapps/contact/contact
miniapps/contact/ParaView
miniapps/plasma/pic/electrostatic-*
!miniapps/plasma/pic/electrostatic-*.cpp
miniapps/plasma/pic/*.csv
# Unit test binary and outputs
tests/unit/output_meshes
tests/unit/unit_tests
-5
View File
@@ -85,8 +85,3 @@ opt_par_gcc_10_pumi:
extends: .mfem_job_on_dane
variables:
SPEC: "%gcc@10.3.1 +pumi"
opt_par_gcc_10_gslib:
extends: .mfem_job_on_dane
variables:
SPEC: "%gcc@10.3.1 +gslib"
-5
View File
@@ -63,8 +63,3 @@ opt_mpi_cuda_hypre_cuda_gcc:
extends: .mfem_job_on_matrix
variables:
SPEC: "%gcc@10.3.1 +mpi +cuda cuda_arch=90 ^hypre+cuda"
opt_mpi_cuda_gcc_gslib:
extends: .mfem_job_on_matrix
variables:
SPEC: "%gcc@10.3.1 +mpi +cuda +gslib cuda_arch=90 ^hypre+cuda"
+2 -2
View File
@@ -32,9 +32,9 @@ mkdir _${BASELINE_TEST} && cd _${BASELINE_TEST}
# run
if [[ "${MACHINE_NAME}" == "dane" ]]; then
srun --nodes=1 -t 60 --exclusive --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
salloc --nodes=1 -t 60 --exclusive --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
elif [[ ${MACHINE_NAME} == "corona" ]]; then
srun --nodes=1 -t 60 -p pbatch ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
salloc --nodes=1 -t 60 -p pbatch ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
else
echo "Unknown machine: MACHINE_NAME=$MACHINE_NAME"
exit 1
-98
View File
@@ -11,87 +11,6 @@
Version 4.9.1 (development)
===========================
- Added policy for AI-assisted contribution to CONTRIBUTING.md.
Discretization improvements
---------------------------
- Improved FindPointsGSLIB surface mesh capability with support for simplices
and an option to specify axis-aligned bounding box padding for near-surface
point queries.
- Added GPU-enabled partial assembly for simplicial Bernstein H1 basis based on
ragged tensor algorithms (see DOI: 10.1137/11082539X) for mass and diffusion
integrators.
- Replaced legacy simplex quadrature rules with symmetric positive weight rules
for triangles (orders 0-25) and tetrahedra (orders 0-20). These rules
guarantee all-positive weights and interior quadrature points, improving
numerical stability. Higher orders fall back to Grundmann-Moller.
* Triangle rules: Witherden and Vincent, DOI: 10.1016/j.camwa.2015.03.017
* Tet rules (d=1-13): Witherden and Vincent (same as above)
* Tet rules (d=14-20): Chuluunbaatar et al., DOI: 10.1016/j.camwa.2022.08.016
- Added support for general 1D Gauss-Jacobi quadrature rules and Stroud conical
quadrature rules on triangles and tetrahedra.
- Improved the GridFunction projection routines. Projections work for Scalar,
Vector and VectorFE, also NURBS versions. Optionally different types of
projections can be selected, default behavior has not changed.
- Added GridFunction projection methods for trace spaces, i.e., project
coefficients on the mesh skeleton.
- Added methods to estimate function extremum using piecewise linear bounds plus
recursive subdivision.
- Extend FindPointsGSLIB to support surface meshes.
Meshing improvements
--------------------
- Added option to guarantee mesh validity during TMOP-based r-adaptivity, using
bounds on the determinant of the mesh transformation Jacobian.
- Added PA support for TMOP's adaptive limiting functionality. Multiple
GridFunctions and Coefficients can be combined to form a composite term.
- Improved support for 1D NURBS meshes with variable order, including using
the patches construct for 1D NURBS meshes.
- Added the option to include material interfaces (faces separating elements
with different element attributes) as additional boundary elements, for
parallel visualization, e.g. with GLVis. This is supported by both the Print
and PrintAsOne methods of ParMesh. See ParMesh::SetPrintInterfaces().
Linear and nonlinear solvers
----------------------------
- Added support for trace spaces in PRefinementTransferOperator. This is used in
PRefinement multigrid methods for problems posed on trace spaces (see e.g. the
DPG miniapps).
GPU computing
-------------
- Added device assembly support for 3D H(curl) VectorFEDomainLFIntegrator.
- Added NVIDIA cuDSS library interface. Implementation examples have been
added to ex1 and ex1p. See https://developer.nvidia.com/cudss for more
details. Supported versions >= 0.6.0.
- Allow specifying GPU kernel launch bounds for native and RAJA GPU backends.
New and updated examples and miniapps
-------------------------------------
- The Lorentz miniapp (in miniapps/electromagnetics) has been updated to
leverage the ParticleSet capability.
- Added (Complex)PRefinementMultigrid solver option in the DPG miniapps.
Miscellaneous
-------------
- Fixed signed DOF handling in ParGridFunction reading (read constructor) and
saving via SaveAsOne(). Simplified the process of applying the DOF signs by
using the new method ApplyDofSigns() in class ParFiniteElementSpace: the
method will return immediately if no sign flips are needed.
Version 4.9, released on Dec 11, 2025
=====================================
@@ -176,23 +95,6 @@ Linear and nonlinear solvers
Filtering (AMGF), providing robust preconditioning for linear systems arising
in constrained optimization problems such as frictionless contact.
Added 'GetResiduals' and 'GetFinalAbsResidualNorm' to 'HyprePCG',
'HypreGMRES', and 'HypreFGMRES' to get 'r' and '|r|_p'. Note that the latter
computes '|r|_p' from 'r' instead of returning a cached value like the
relative 'GetFinalResidualNorm'. These require Hypre >= 2.15.0.
Changed the default solver parameters for 'HyprePCG' to 'tol=1e-6' and
'max_iter=1000'. This matches the default parameters in Hypre 3.0.
Added various helper functions for querying/modifying Hypre solvers:
'HypreSmoother::GetType', 'HypreSmoother::GetSOROptions',
'HypreSmoother::GetPolyOptions', 'HypreSmoother::GetWindowParameters',
'HypreSmoother::IsOperatorSymmetric', 'HyprePCG::GetTol',
'HyprePCG::GetAbsTol', 'HyprePCG::GetMaxIter', 'HyprePCG::SetUseTwoNorm',
'HypreGMRES::GetTol', 'HypreGMRES::GetAbsTol', 'HypreGMRES::GetMaxIter',
'HypreGMRES::GetKDim', 'HypreFGMRES::GetTol', 'HypreFGMRES::GetMaxIter',
'HypreFGMRES::GetKDim', and 'HypreBoomerAMG::GetMaxIter'.
GPU computing
-------------
- Added the 'gpu', 'raja-gpu', and 'ceed-gpu' backend aliases/shortcuts which
+7 -25
View File
@@ -433,15 +433,6 @@ if (MFEM_USE_STRUMPACK)
endif()
endif()
# cuDSS can only be enabled in CUDA
if (MFEM_USE_CUDSS)
if (MFEM_USE_CUDA)
find_package(CUDSS REQUIRED)
else()
message(FATAL_ERROR " *** cuDSS requires that CUDA be enabled.")
endif()
endif()
# GnuTLS
if (MFEM_USE_GNUTLS)
find_package(_GnuTLS REQUIRED)
@@ -640,7 +631,7 @@ find_package(Threads REQUIRED)
set(MFEM_TPLS OPENMP HYPRE LAPACK BLAS SuperLUDist STRUMPACK METIS SuiteSparse
SUNDIALS PETSC SLEPC MUMPS AXOM FMS CONDUIT Ginkgo GNUTLS GSLIB HDF5
NETCDF MPFR PUMI HIOP POSIXCLOCKS MFEMBacktrace ZLIB OCCA CEED RAJA UMPIRE
ADIOS2 MKL_CPARDISO MKL_PARDISO AMGX MAGMA CUSPARSE CUBLAS CUDSS CALIPER CODIPACK
ADIOS2 MKL_CPARDISO MKL_PARDISO AMGX MAGMA CUSPARSE CUBLAS CALIPER CODIPACK
BENCHMARK PARELAG TRIBOL MPI_CXX HIP HIPBLAS HIPSPARSE MOONOLITH BLITZ
ALGOIM ENZYME CUDA::cudart)
@@ -661,8 +652,6 @@ foreach(TPL IN LISTS MFEM_TPLS)
endif()
endforeach(TPL)
# reverse to remove the first instance of entries in TPL_LIBRARIES
# so later duplicates are kept (for dependency ordering)
list(REVERSE TPL_LIBRARIES)
list(REMOVE_DUPLICATES TPL_LIBRARIES)
list(REVERSE TPL_LIBRARIES)
@@ -881,12 +870,11 @@ add_dependencies(exec
# - https://cmake.org/Bug/view.php?id=8438
# Add a target to copy the mfem data directory to the build directory
# Implementable as a single copy_directory_if_different command w/ CMake >= 3.26
file(GLOB DATA_FILES CONFIGURE_DEPENDS ${PROJECT_SOURCE_DIR}/data/*)
add_custom_target(copy_data
COMMAND ${CMAKE_COMMAND} -E make_directory data
COMMAND ${CMAKE_COMMAND} -E copy_if_different ${DATA_FILES} data
COMMENT "Syncing the data directory ...")
add_custom_command(OUTPUT data_is_copied
COMMAND ${CMAKE_COMMAND} -E copy_directory ${PROJECT_SOURCE_DIR}/data data
COMMAND ${CMAKE_COMMAND} -E touch data_is_copied
COMMENT "Copying the data directory ...")
add_custom_target(copy_data DEPENDS data_is_copied)
# Add 'copy_data' as a prerequisite for all executables, if the source and the
# build directories are not the same.
if (NOT ("${PROJECT_SOURCE_DIR}" STREQUAL "${PROJECT_BINARY_DIR}"))
@@ -1018,15 +1006,9 @@ install(FILES
install(EXPORT ${PROJECT_NAME_UC}Targets
DESTINATION ${INSTALL_CMAKE_DIR})
# Install the data directory if present, i.e. if the copy_data target is built
install(DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/data
DESTINATION ${MFEM_INSTALL_DIR} OPTIONAL)
#-------------------------------------------------------------------------------
# Create 'config.mk' from 'config.mk.in' for the build and install locations and
# define install rules for 'config.mk' and 'test.mk'
#-------------------------------------------------------------------------------
if (MFEM_USE_CUDA OR MFEM_USE_HIP)
option(MFEM_EXPORT_GPU_CONFIG "Export config.mk for GPU-enabled downstream packages" ON)
endif()
mfem_export_mk_files()
+65 -72
View File
@@ -3,12 +3,12 @@
</p>
<p align="center">
<a href="https://github.com/mfem/mfem/blob/master/LICENSE"><img alt="License" src="https://img.shields.io/badge/License-BSD-blue.svg"></a>
<a href="https://github.com/mfem/mfem/releases/latest"><img alt="GitHub release" src="https://img.shields.io/github/v/release/mfem/mfem"></a>
<a href="https://github.com/mfem/mfem/actions/workflows/repo-check.yml?query=branch%3Amaster"><img alt="Repo check" src="https://github.com/mfem/mfem/actions/workflows/repo-check.yml/badge.svg?branch=master"></a>
<a href="https://github.com/mfem/mfem/actions/workflows/builds-and-tests.yml?query=branch%3Amaster"><img alt="Builds and Tests" src="https://github.com/mfem/mfem/actions/workflows/builds-and-tests.yml/badge.svg?branch=master"></a>
<a href="https://github.com/mfem/mfem/blob/master/LICENSE"><img alt="License" src="https://img.shields.io/badge/License-BSD-brightgreen.svg"></a>
<a href="https://github.com/mfem/mfem/actions?query=workflow%3Arepo-check+branch%3Amaster"><img alt="Repo check" src="https://github.com/mfem/mfem/actions/workflows/repo-check.yml/badge.svg?branch=master"></a>
<a href="https://github.com/mfem/mfem/actions?query=workflow%3Abuild-analysis+branch%3Amaster"><img alt="Build Analysis" src="https://github.com/mfem/mfem/actions/workflows/mfem-analysis.yml/badge.svg?branch=master"></a>
<a href="https://github.com/mfem/mfem/actions?query=workflow%3Abuilds-and-tests+branch%3Amaster"><img alt="Builds and Tests" src="https://github.com/mfem/mfem/actions/workflows/builds-and-tests.yml/badge.svg?branch=master"></a>
<a href="https://ci.appveyor.com/project/mfem/mfem"><img alt="Build Status" src="https://ci.appveyor.com/api/projects/status/19non9sqm6msi2wy?svg=true"></a>
<a href="https://docs.mfem.org/html/index.html"><img alt="Documentation" src="https://img.shields.io/badge/code-documented-brightgreen.svg"></a>
<a href="https://docs.mfem.org/html/index.html"><img alt="Doxygen" src="https://img.shields.io/badge/code-documented-brightgreen.svg"></a>
</p>
@@ -24,14 +24,6 @@ must be made under this license.
Note also that MFEM has a [Code of Conduct](CODE_OF_CONDUCT.md). By participating
in the MFEM community, you agree to abide by its rules.
## AI Policy
- Use of AI code generation in MFEM is allowed but must be disclosed, e.g. by
selecting the `AI-assisted` label on the PR.
- By submitting a PR, the author acknowledges that they have reviewed and
understand the changes they are proposing.
- PR authors are still responsible for correctness, licensing, and attribution
of all changes.
If you plan on contributing to MFEM, consider reviewing the
[issue tracker](https://github.com/mfem/mfem/issues) first to check if a thread
already exists for your desired feature or the bug you ran into. Use a pull
@@ -84,7 +76,7 @@ Origin](#developers-certificate-of-origin-11) at the end of this file.*
follow the [MFEM PR Rules](#mfem-pr-rules).
- When your contribution is fully working and ready to be reviewed, add
the `ready-for-review` label.
- PRs are treated similarly to journal submission, with an "editor" assigning two
- PRs are treated similarly to journal submission with an "editor" assigning two
reviewers to evaluate the changes.
- The reviewers have 3 weeks to evaluate the PR and work with the author to
fix issues and implement improvements.
@@ -125,7 +117,7 @@ The MFEM source code has the following structure:
│ ├── petsc
│ ├── pumi
│ ├── sundials
└── superlu
| └── superlu
├── fem
│ ├── ceed
│ ├── dfem
@@ -137,6 +129,10 @@ The MFEM source code has the following structure:
│ ├── moonolith
│ ├── qinterp
│ └── tmop
│ | ├── assemble
│ | ├── metrics
│ | ├── mult
│ | └── tools
├── general
├── linalg
│ ├── batched
@@ -149,10 +145,11 @@ The MFEM source code has the following structure:
│ ├── common
│ ├── contact
│ ├── dfem
│ ├── diag-smoothers
│ ├── dpg
│ ├── electromagnetics
│ ├── fluids
│ │ ├── navier
│ │ └── schrodinger-flow
│ ├── gslib
│ ├── hdiv-linear-solver
│ ├── hooke
@@ -162,7 +159,6 @@ The MFEM source code has the following structure:
│ ├── nurbs
│ ├── parelag
│ ├── performance
│ ├── plasma
│ ├── shifted
│ ├── solvers
│ ├── spde
@@ -193,15 +189,15 @@ respectively.
- The main finite element classes are:
+ [`FiniteElement`](https://docs.mfem.org/html/classmfem_1_1FiniteElement.html)
+ [`FiniteElementCollection`](https://docs.mfem.org/html/classmfem_1_1FiniteElementCollection.html)
+ [`FiniteElementCollection`](https://docs.mfem.org/html/classmfem_1_1FiniteElement.html)
+ [`FiniteElementSpace`](https://docs.mfem.org/html/classmfem_1_1FiniteElementSpace.html)
+ [`GridFunction`](https://docs.mfem.org/html/classmfem_1_1GridFunction.html)
+ [`BilinearFormIntegrator`](https://docs.mfem.org/html/classmfem_1_1BilinearFormIntegrator.html) and [`LinearFormIntegrator`](https://docs.mfem.org/html/classmfem_1_1LinearFormIntegrator.html)
+ [`LinearForm`](https://docs.mfem.org/html/classmfem_1_1LinearForm.html), [`BilinearForm`](https://docs.mfem.org/html/classmfem_1_1BilinearForm.html) and [`MixedBilinearForm`](https://docs.mfem.org/html/classmfem_1_1MixedBilinearForm.html)
+ [`LinearForm`](https://docs.mfem.org/html/classmfem_1_1LinearFormIntegrator.html), [`BilinearForm`](https://docs.mfem.org/html/classmfem_1_1BilinearForm.html) and [`MixedBilinearForm`](https://docs.mfem.org/html/classmfem_1_1MixedBilinearForm.html)
- The main linear algebra classes and sources are
+ [`Operator`](https://docs.mfem.org/html/classmfem_1_1Operator.html) and [`BilinearForm`](https://docs.mfem.org/html/classmfem_1_1BilinearForm.html)
+ [`Vector`](https://docs.mfem.org/html/classmfem_1_1Vector.html) and [`LinearForm`](https://docs.mfem.org/html/classmfem_1_1LinearForm.html)
+ [`Vector`](https://docs.mfem.org/html/classmfem_1_1BilinearForm.html) and [`LinearForm`](https://docs.mfem.org/html/classmfem_1_1LinearForm.html)
+ [`DenseMatrix`](https://docs.mfem.org/html/classmfem_1_1DenseMatrix.html) and [`SparseMatrix`](https://docs.mfem.org/html/classmfem_1_1SparseMatrix.html)
+ Sparse [smoothers](https://docs.mfem.org/html/sparsesmoothers_8hpp.html) and linear [solvers](https://docs.mfem.org/html/solvers_8hpp.html)
@@ -213,8 +209,8 @@ shared geometric entities between different tasks. The parallel source files
have a `p` prefix, e.g. `pmesh.cpp` vs. the serial `mesh.cpp`.
- The main parallel classes are
+ [`ParMesh`](https://docs.mfem.org/html/classmfem_1_1ParMesh.html)
+ [`ParNCMesh`](https://docs.mfem.org/html/classmfem_1_1ParNCMesh.html)
+ [`ParMesh`](https://docs.mfem.org/html/solvers_8hpp.html)
+ [`ParNCMesh`](https://docs.mfem.org/html/classmfem_1_1ParMesh.html)
+ [`ParFiniteElementSpace`](https://docs.mfem.org/html/classmfem_1_1ParFiniteElementSpace.html)
+ [`ParGridFunction`](https://docs.mfem.org/html/classmfem_1_1ParGridFunction.html)
+ [`ParBilinearForm`](https://docs.mfem.org/html/classmfem_1_1ParBilinearForm.html) and [`ParLinearForm`](https://docs.mfem.org/html/classmfem_1_1ParLinearForm.html)
@@ -224,14 +220,14 @@ have a `p` prefix, e.g. `pmesh.cpp` vs. the serial `mesh.cpp`.
#### GPU and general device support
GPU and multi-core CPU support is based on device kernels supporting different
backends (CUDA, HIP, OCCA, RAJA, OpenMP, etc.) and an internal lightweight
backends (CUDA, OCCA, RAJA, OpenMP, etc.) and an internal lightweight
device/host memory manager.
- The main device-relevant classes and sources are:
+ [`Device`](https://docs.mfem.org/html/device_8hpp.html)
+ [`MemoryManager`](https://docs.mfem.org/html/mem_manager_8hpp.html)
+ the [`mfem::forall`](https://docs.mfem.org/html/forall_8hpp.html) function
+ the [`cuda.hpp`](https://docs.mfem.org/html/cuda_8hpp.html), [`hip.hpp`](https://docs.mfem.org/html/hip_8hpp.html) and [`occa.hpp`](https://docs.mfem.org/html/occa_8hpp.html) files
+ the [`cuda.hpp`](https://docs.mfem.org/html/cuda_8hpp.html) and [`occa.hpp`](https://docs.mfem.org/html/occa_8hpp.html) files
#### Utilities, building and documentation
- The `general/` directory contains C++ classes that serve as utilities for
@@ -245,8 +241,8 @@ device/host memory manager.
- `examples` and `miniapps` respectively gather simple and more fully-featured
demonstrations of the usage on MFEM. They both rely on `data/` for the
collection of meshes.
- The `tests/` directory contains a unit test suite, additional tests, and
benchmarks.
- The `tests/` directory contains a unit test suite and will later contain more
tests that run example codes.
See also the [code overview](https://mfem.org/code-overview/) section on the MFEM
website.
@@ -280,8 +276,8 @@ Before you can start, you need a GitHub account, here are a few suggestions:
the top of https://github.com/mfem.
- Consider making your membership public by going to https://github.com/orgs/mfem/people
and clicking on the organization visibility drop box next to your name.
- Project discussions and announcements will be posted at https://github.com/orgs/mfem/discussions,
tagging the `@mfem/everyone` team when appropriate.
- Project discussions and announcements will be posted at
https://github.com/orgs/mfem/teams/everyone.
#### Structure
- The MFEM source code is in the [mfem](https://github.com/mfem/mfem)
@@ -341,12 +337,11 @@ Before you can start, you need a GitHub account, here are a few suggestions:
- Well-designed simple code is frequently more general and powerful.
- Lean code base is easier to understand by new collaborators.
- New features should be added only if they are necessary or generally useful.
- Introduction of language constructs not currently used in MFEM should be
- Introduction of language constructions not currently used in MFEM should be
justified and generally avoided (to maintain portability to various systems
and compilers, including early access hardware).
- We prefer basic C++. Use C++17 features judiciously, prioritizing readability,
consistency with existing MFEM code, and portability to different systems,
compilers and device backends.
- We prefer basic C++ and the C++03 standard, to keep the code readable by
a large audience and to make sure it compiles anywhere.
- *Keep the code general and reasonably efficient*
- The main goal is fast prototyping for research and application development.
@@ -389,7 +384,7 @@ Before you can start, you need a GitHub account, here are a few suggestions:
- When your branch is ready for other developers to review / comment on
the code, create a pull request towards `mfem:master`.
- Pull requests typically have titles like:
- Pull request typically have titles like:
`Description [new-feature-dev]`
@@ -410,12 +405,12 @@ Before you can start, you need a GitHub account, here are a few suggestions:
- Add a description, appropriate labels and assign yourself to the PR. The MFEM
team will add reviewers as appropriate.
- List outstanding TODO items in the description.
- List outstanding TODO items in the description, see PR #222 for an example.
- When your contribution is fully working and ready to be reviewed, add
or request the `ready-for-review` label.
the `ready-for-review` label.
- PRs are treated similarly to journal submission, with an "editor" assigning
- PRs are treated similarly to journal submission with an "editor" assigning
two reviewers to evaluate the changes. The reviewers have 3 weeks to evaluate
the PR and work with the author to implement improvements and fix issues.
@@ -441,7 +436,7 @@ Before you can start, you need a GitHub account, here are a few suggestions:
checks in GitHub Actions enforce MFEM-specific rules which are explained in
the error messages and the `tests/scripts` directory.
- Also note that the tests `branch-history` and `repo-check` found in GitHub
- Also note that the tests `branch-history` and `repos-checks` found in GitHub
Actions can be triggered automatically before each push using git hooks. See
the [git hooks README](config/githooks/README.md) for a detailed explanation.
@@ -498,15 +493,15 @@ Everyone on the MFEM team can be asked to serve as a reviewer on a PR in their a
3. To ensure the quality of the PR by making sure that the code adheres to the [Developer Guidelines](#developer-guidelines), e.g. all methods, data members, and functions have documentation, including data ownership and lifetime, new examples/miniapps have a corresponding PR in mfem/web, major features have `CHANGELOG` entries, etc.
4. To seek help from the editors in case of difficulties.
3. To seek help from the editors in case of difficulties.
5. To complete the review in a timely manner: 3 weeks from assignment.
4. To complete the review in a timely manner: 3 weeks from assignment.
6. To test the PR thoroughly before merging in *next*. The PR author is also encouraged to perform testing and inform the reviewers about the results.
5. To test the PR thoroughly before merging in *next*. The PR author is also encouraged to perform testing and inform the reviewers about the results.
7. To monitor the PR impact on the testing in the *next* branch and alert the editors that the PR is ready for merging in *master*.
6. To monitor the PR impact on the testing in the *next* branch and alert the editors that the PR is ready for merging in *master*.
8. The review of bugfixes should be expedited proportional to their importance. The review window can be much less than three weeks in such cases.
7. The review of bugfixes should be expedited proportional to their importance. The review window can be much less than three weeks in such cases.
#### Responsibilities of Authors
@@ -532,30 +527,30 @@ Before a PR can be merged, it should satisfy the following:
- [ ] Code builds.
- [ ] Code passes `make style`.
- [ ] Update `CHANGELOG`:
- [ ] Is this a new feature users need to be aware of? New or updated example or miniapp?
- [ ] Does it make sense to create a new section in the `CHANGELOG` to group with other related features?
- [ ] Is this a new feature users need to be aware of? New or updated example or miniapp?
- [ ] Does it make sense to create a new section in the `CHANGELOG` to group with other related features?
- [ ] Update `INSTALL`:
- [ ] Has a new optional library been added? If so, what range of versions of this library are required? (*Make sure the external library is compatible with our BSD license, e.g. it is not licensed under GPL!*)
- [ ] Have the version ranges for any required or optional libraries changed?
- [ ] Does `make` or `cmake` have a new target?
- [ ] Did the requirements or the installation process change? *(rare)*
- [ ] Had a new optional library been added? If so, what range of versions of this library are required? (*Make sure the external library is compatible with our BSD license, e.g. it is not licensed under GPL!*)
- [ ] Have the version ranges for any required or optional libraries changed?
- [ ] Does `make` or `cmake` have a new target?
- [ ] Did the requirements or the installation process change? *(rare)*
- [ ] Update continuous integration server configurations if necessary (e.g. with new version requirements for each of MFEM's dependencies)
- [ ] `.github`
- [ ] `.appveyor.yml`
- [ ] `.github`
- [ ] `.appveyor.yml`
- [ ] Update `.gitignore`:
- [ ] Check if `make distclean; git status` shows any files that were generated from the source by the project (not an IDE) but we don't want to track in the repository.
- [ ] Add new patterns (just for the new files above) and re-run the above test.
- [ ] Check if `make distclean; git status` shows any files that were generated from the source by the project (not an IDE) but we don't want to track in the repository.
- [ ] Add new patterns (just for the new files above) and re-run the above test.
- [ ] New examples:
- [ ] All sample runs at the top of the example source file work.
- [ ] Update `examples/makefile`:
- [ ] All sample runs at the top of the example source file work.
- [ ] Update `examples/makefile`:
- [ ] Add the example code to the appropriate `SEQ_EXAMPLES` and `PAR_EXAMPLES` variables.
- [ ] Add any files generated by it to the `clean` target.
- [ ] Add the example binary and any files generated by it to the top-level `.gitignore` file.
- [ ] Update `examples/CMakeLists.txt`:
- [ ] Update `examples/CMakeLists.txt`:
- [ ] Add the example code to the `ALL_EXE_SRCS` variable.
- [ ] Make sure `THIS_TEST_OPTIONS` is set correctly for the new example.
- [ ] List the new example in `doc/CodeDocumentation.dox`.
- [ ] If new examples directory (e.g. `examples/pumi`), list it in `doc/CodeDocumentation.conf.in`
- [ ] If new examples directory (e.g.`examples/pumi`), list it in `doc/CodeDocumentation.conf.in`
- [ ] Companion pull request for documentation in [mfem/web](https://github.com/mfem/web) repo:
- [ ] Update or add example-specific documentation, see e.g. the `src/examples.md`.
- [ ] Add the description, labels and screenshots in `src/examples.md` and `src/img`.
@@ -572,13 +567,13 @@ Before a PR can be merged, it should satisfy the following:
- [ ] Add/update the `CMakeLists.txt` file in the new miniapp directory.
- [ ] Consider adding a new test for the new miniapp.
- [ ] List the new miniapp in `doc/CodeDocumentation.dox`
- [ ] If new miniapps directory (e.g. `miniapps/nurbs`), add it to `MINIAPP_SUBDIRS` in the `makefile`.
- [ ] If new miniapps directory (e.g. `miniapps/nurbs`), list it in `doc/CodeDocumentation.conf.in`
- [ ] If new miniapps directory (e.g.`miniapps/nurbs`), add it to `MINIAPP_SUBDIRS` in the `makefile`.
- [ ] If new miniapps directory (e.g.`miniapps/nurbs`), list it in `doc/CodeDocumentation.conf.in`
- [ ] Companion pull request for documentation in [mfem/web](https://github.com/mfem/web) repo:
- [ ] Update or add miniapp-specific documentation, see e.g. the `src/meshing.md` and `src/electromagnetics.md` files.
- [ ] Add the description, labels and screenshots in `src/examples.md` and `src/img`.
- [ ] The miniapps go at the end of the page, and are usually listed only under a specific "Application (PDE)" category.
- [ ] Add a short description of the miniapp in the "Extensive Examples" section of `features.md`.
- [ ] Update or add miniapp-specific documentation, see e.g. the `src/meshing.md` and `src/electromagnetics.md` files.
- [ ] Add the description, labels and screenshots in `src/examples.md` and `src/img`.
- [ ] The miniapps go at the end of the page, and are usually listed only under a specific "Application (PDE)" category.
- [ ] Add a short description of the miniapp in the "Extensive Examples" section of `features.md`.
- [ ] New capability:
- [ ] All new public, protected, and private classes, methods, data members, and functions have full Doxygen-style documentation in source comments. Documentation should include descriptions of member data, function arguments and return values, template parameters, and prerequisites for calling new functions.
- [ ] Pointer arguments and return values must specify whether ownership is being transferred or lent with the call.
@@ -680,7 +675,7 @@ MFEM uses a `master`/`next`-branch workflow as described below:
- [ ] Update URL shortlinks:
- [ ] Create a shortlink at [http://bit.ly/](http://bit.ly/) for the release tarball, e.g. https://mfem.github.io/releases/mfem-3.1.tgz.
- [ ] (LLNL only) Add and commit the new shortlink in the `links` and `links-mfem` files of the internal `mfem/downloads` repo.
- [ ] Add the new shortlinks to the MFEM package in `spack`.
- [ ] Add the new shortlinks to the MFEM packages in `spack`, `homebrew/science`, `VisIt`, etc.
- [ ] Update website in `mfem/web` repo:
- Update version and shortlinks in `src/index.md` and `src/download.md`.
- Use [cloc-1.62.pl](http://cloc.sourceforge.net/) and `ls -lh` to estimate the SLOC and the tarball size in `src/download.md`.
@@ -732,24 +727,22 @@ commit or push, see the [README](config/githooks/README.md) in the `config/githo
directory.
### GitHub Actions smoke tests
### Linux and Mac smoke tests
We use GitHub Actions to drive the default tests on the `master` and `next`
branches. See the `.github/workflows` files and the logs at
[https://github.com/mfem/mfem/actions](https://github.com/mfem/mfem/actions).
GitHub Actions testing should be kept lightweight, as there is a time
constraint on jobs. The current workflows cover Linux, macOS, and Windows
configurations.
Testing using GitHub Actions should be kept lightweight, as there is a time
constraint on jobs. Two virtual machines are configured - Mac (OS X) and Linux.
- Tests on the `master` branch are triggered whenever a PR is issued on this branch.
- Tests on the `next` branch are currently scheduled to run each night.
### Additional Windows smoke test
We also use Appveyor to test building with the MS Visual C++ compiler in a Windows
environment, as well as to test the CMake build. See the `.appveyor.yml` file
and the build logs at
### Windows smoke test
We use Appveyor to test building with the MS Visual C++ compiler in a Windows
environment, as well as to test the CMake build. See the `.appveyor` file and the
build logs at
[https://ci.appveyor.com/project/mfem/mfem](https://ci.appveyor.com/project/mfem/mfem).
CMake is used to generate the MSVC Project files and drive the build. A release
+16 -31
View File
@@ -38,13 +38,14 @@ the option MFEM_USE_METIS.
MFEM also includes support for devices such as GPUs, and programming models such
as CUDA, HIP, OCCA, OpenMP and RAJA.
- Starting with version 4.9, MFEM requires a C++17 compiler.
- Starting with version 4.0, MFEM requires a C++11 compiler. We recommend using
a newer compiler, e.g. GCC version 4.9 or higher.
- CUDA support requires an NVIDIA GPU and an installation of the CUDA Toolkit
https://developer.nvidia.com/cuda-toolkit
- HIP support requires an AMD GPU and an installation of the ROCm software stack
https://rocm.docs.amd.com
https://rocmdocs.amd.com
- OCCA support requires the OCCA library
https://libocca.org
@@ -82,9 +83,9 @@ Serial build:
Parallel build:
(download hypre and METIS 4 from above URLs)
(build METIS 4 in ../metis-4.0 relative to mfem/)
(for METIS 5, see https://mfem.org/building/#parallel-build-using-metis-5)
(build hypre in ../hypre relative to mfem/)
make parallel -j 4
(For METIS 5, see https://mfem.org/building/#parallel-build-using-metis-5)
CUDA build:
make cuda -j 4
@@ -114,14 +115,14 @@ Serial build:
Parallel build:
(download hypre and METIS 4 from above URLs)
(build METIS 4 in ../metis-4.0 relative to mfem/)
(for METIS 5, see https://mfem.org/building/#parallel-build-using-metis-5)
(build hypre in ../hypre relative to mfem/)
mkdir <mfem-build-dir> ; cd <mfem-build-dir>
cmake <mfem-source-dir> -DMFEM_USE_MPI=YES
make -j 4
(For METIS 5, see https://mfem.org/building/#parallel-build-using-metis-5)
Parallel build with fetching of hypre and METIS:
mkdir <mfem-build-dir> ; cd <mfem-build-dir>
mkdir <mfem-buil-dir> ; cd <mfem-build-dir>
cmake <mfem-source-dir> -DMFEM_USE_MPI=YES -DMFEM_FETCH_TPLS=YES
make -j 4
@@ -133,8 +134,7 @@ CUDA build:
HIP build:
mkdir <mfem-build-dir> ; cd <mfem-build-dir>
cmake <mfem-source-dir> -DMFEM_USE_HIP=YES -DHIP_ARCH=gfx942 \
-DCMAKE_CXX_COMPILER=amdclang++ -DCMAKE_HIP_COMPILER=amdclang++
cmake <mfem-source-dir> -DMFEM_USE_HIP=YES -DHIP_ARCH=gfx942 -DCMAKE_CXX_COMPILER=amdclang++ -DCMAKE_HIP_COMPILER=amdclang++
make -j 4
Example codes (serial/parallel, depending on the build):
@@ -269,7 +269,6 @@ Compilers:
CXX - C++ compiler, serial build
MPICXX - MPI C++ compiler, parallel build
CUDA_CXX - The CUDA compiler, 'nvcc' or 'clang++'
HIP_CXX - The HIP compiler, e.g. 'hipcc'
Compiler options:
OPTIM_FLAGS - Options for optimized build
@@ -396,11 +395,6 @@ MFEM_USE_STRUMPACK = YES/NO
classes. When enabled, this option uses the STRUMPACK_* library options, see
below.
MFEM_USE_CUDSS = YES/NO
Enable MFEM functionality based on the cuDSS library. When using cuDSS, CUDA
support must be also enabled in MFEM, i.e. MFEM_USE_CUDA=YES must be set.
When enabled, this option uses the CUDSS_* library options, see below.
MFEM_USE_GINKGO = YES/NO
Enable MFEM functionality based on the Ginkgo library, which provides
iterative linear solvers and preconditioners with OpenMP, CUDA backends, see
@@ -560,13 +554,13 @@ MFEM_USE_RAJA = YES/NO
MFEM_USE_OCCA = YES/NO
Enables support for the OCCA library in MFEM. OCCA is an open-source library
which aims to make it easy to program different types of devices (e.g. CPU,
GPU, FPGA) by providing a unified API for interacting with JIT-compiled
GPU, FPGA) by providing an unified API for interacting with JIT-compiled
backends. In order to use the OCCA CUDA backend, CUDA support must be enabled
in MFEM as well, i.e. MFEM_USE_CUDA=YES must be set.
MFEM_USE_GSLIB = YES/NO
Enables MFEM functionality based on the GSLIB library, and specifically its
FindPoints component, which provides robust algorithms to evaluate finite
FindPoints component, which provides a robust algorithms to evaluate finite
element functions in a collection of points in physical space. When enabled,
the user can use the GSLIB-FindPoints methods as shown in miniapps/gslib.
@@ -725,18 +719,9 @@ The specific libraries and their options are:
Options: STRUMPACK_OPT, STRUMPACK_LIB.
Versions: STRUMPACK >= 3.0.0.
- CUDSS (optional), used when MFEM_USE_CUDSS = YES. Note that CUDSS requires
CUDA 12.x toolkit and the cuDSS libraries. The supported communication backend
is OpenMPI 4.x (default), and OpenMPI 4.x or a later version must be pre-built.
The source files in the cuDSS tarball provide guidance for developing custom
MPI implementations.
URL: https://developer.nvidia.com/cudss
https://docs.nvidia.com/cuda/cudss/advanced_features.html#communication-layer-library-in-cudss
Options: CUDSS_OPT, CUDSS_LIB.
Versions: cuDSS >= 0.6.0.
- Ginkgo (optional), used when MFEM_USE_GINKGO = YES. Ginkgo may have additional
requirements and module-specific dependencies; see the webpage below.
- Ginkgo (optional), used when MFEM_USE_GINKGO = YES. Note that Ginkgo needs a
C++ compiler that supports the C++-17 standard. For additional requirements
and dependencies of specific modules, see the Ginkgo webpage below.
URL: https://ginkgo-project.github.io
Options: GINKGO_OPT, GINKGO_LIB, GINKGO_DIR, GINKGO_BUILD_TYPE (Release or
Debug).
@@ -808,7 +793,7 @@ The specific libraries and their options are:
Options: CONDUIT_OPT, CONDUIT_LIB.
Versions: Conduit >= 0.3.1.
- ADIOS2 (optional), used when MFEM_USE_ADIOS2 = YES.
- ADIOS2 (optional) used when MFEM_USE_ADIOS2 = YES.
URL: https://adios2.readthedocs.io/
Versions: ADIOS >= 2.5.0.
@@ -884,7 +869,7 @@ The specific libraries and their options are:
Options: RAJA_DIR, RAJA_OPT, RAJA_LIB.
Versions: RAJA >= 2022.10.3.
- Moonolith (optional), used when MFEM_USE_MOONOLITH = YES.
- Moonolith (optional), use when MFEM_USE_MOONOLITH = YES.
URL: https://bitbucket.org/zulianp/par_moonolith
Options: MOONOLITH_DIR
Versions: MOONOLITH >= 1.1.0.
@@ -972,7 +957,7 @@ CMAKE_BUILD_TYPE which can be set to standard values like "Debug", and "Release"
To use a specific generator use the "-G <generator>" option of cmake:
cmake <mfem-source-dir> -G "Xcode"
cmake <mfem-source-dir> -G "Visual Studio 17 2022"
cmake <mfem-source-dir> -G "Visual Studio 12 2013"
cmake <mfem-source-dir> -G "MinGW Makefiles"
With CMake it is possible to build MFEM as a shared library using the standard
@@ -1217,7 +1202,7 @@ larger problems, there are two options:
Specific options for HIP
========================
MFEM expects the `ROCM_PATH` environment variable to be set to the path of the
ROCm install, as well as having `$ROCM_PATH/bin` in `PATH`.
ROCM install, as well as having `$ROCM_PATH/bin` in `PATH`.
Specific options for RAJA+HIP+MPI
=================================
-1
View File
@@ -28,7 +28,6 @@ license files. These software products and their licenses are as follows:
* AmgXWrapper (linalg/amgxsolver.{hpp,cpp}) -- MIT license
* Catch++ (tests/unit/catch.hpp) -- Boost 1.0 license
* Gecko (general/gecko.{cpp,hpp}) -- BSD 3-clause license
* gslib (fem/gslib.{cpp,hpp}, mesh/bb_grid_map.{cpp,hpp}) -- BSD 3-clause license
* Picojson (fem/picojson.h) -- Custom 2-clause license
* TinyXML2 (general/tinyxml2.{cpp,h}) -- zlib license
* Zstr (general/zstr.hpp) -- MIT license
-9
View File
@@ -35,7 +35,6 @@ set(MFEM_USE_SUITESPARSE @MFEM_USE_SUITESPARSE@)
set(MFEM_USE_SUPERLU @MFEM_USE_SUPERLU@)
set(MFEM_USE_MUMPS @MFEM_USE_MUMPS@)
set(MFEM_USE_STRUMPACK @MFEM_USE_STRUMPACK@)
set(MFEM_USE_CUDSS @MFEM_USE_CUDSS@)
set(MFEM_USE_GINKGO @MFEM_USE_GINKGO@)
set(MFEM_USE_AMGX @MFEM_USE_AMGX@)
set(MFEM_USE_MAGMA @MFEM_USE_MAGMA@)
@@ -110,14 +109,6 @@ if (MFEM_USE_RAJA)
find_dependency(RAJA)
endif()
if (MFEM_USE_CUDSS)
find_dependency(cudss)
endif (MFEM_USE_CUDSS)
if (MFEM_USE_UMPIRE)
find_dependency(umpire)
endif()
if (NOT TARGET mfem)
include(${CMAKE_CURRENT_LIST_DIR}/MFEMTargets.cmake)
endif (NOT TARGET mfem)
-9
View File
@@ -108,15 +108,6 @@
// Enable MFEM functionality based on the STRUMPACK library.
#cmakedefine MFEM_USE_STRUMPACK
// Enable MFEM functionality based on the cuDSS library.
#cmakedefine MFEM_USE_CUDSS
// CUDSS communication layer library path
#cmakedefine MFEM_CUDSS_COMM_LIB "@MFEM_CUDSS_COMM_LIB@"
// CUDSS threading layer library path
#cmakedefine MFEM_CUDSS_THREADING_LIB "@MFEM_CUDSS_THREADING_LIB@"
// Enable functionality based on the Ginkgo library.
#cmakedefine MFEM_USE_GINKGO
-68
View File
@@ -1,68 +0,0 @@
if (NOT cudss_DIR AND CUDSS_DIR)
set(cudss_DIR ${CUDSS_DIR}/lib/cmake/cudss)
endif()
message(STATUS "Looking for CUDSS ...")
message(STATUS " in CUDSS_DIR = ${CUDSS_DIR}")
message(STATUS " cudss_DIR = ${cudss_DIR}")
find_package(cudss)
set(CUDSS_FOUND ${cudss_FOUND})
set(CUDSS_LIBRARIES "cudss")
if (CUDSS_FOUND)
message(STATUS
"Found CUDSS target: ${CUDSS_LIBRARIES} (version: ${cudss_VERSION})")
else()
set(msg STATUS)
if (CUDSS_FIND_REQUIRED)
set(msg FATAL_ERROR)
endif()
message(${msg}
"CUDSS not found. Please set CUDSS_DIR to the install prefix.")
endif()
if(CUDSS_FOUND AND TARGET cudss)
get_target_property(CUDSS_LIBRARY_LOCATION cudss IMPORTED_LOCATION)
if(NOT CUDSS_LIBRARY_LOCATION)
get_target_property(CUDSS_LIBRARY_LOCATION cudss IMPORTED_LOCATION_RELEASE)
endif()
if(CUDSS_LIBRARY_LOCATION)
get_filename_component(CUDSS_LIBRARY_DIR "${CUDSS_LIBRARY_LOCATION}" DIRECTORY)
else()
message(WARNING "Could not determine the location of the cuDSS library.")
endif()
else()
message(WARNING "cuDSS target not available; cannot determine library directory.")
endif()
# Set the full name of the cuDSS threading library if OpenMP is enabled.
# The threading layer library (libcudss_mtlayer_gomp.so) is located under the
# cuDSS library directory by default.
if (MFEM_USE_OPENMP)
find_file(
CUDSS_THREADING_LIB
NAMES libcudss_mtlayer_gomp.so
PATHS ${CUDSS_LIBRARY_DIR}
NO_DEFAULT_PATH
)
if (NOT DEFINED MFEM_CUDSS_THREADING_LIB AND CUDSS_THREADING_LIB)
set(MFEM_CUDSS_THREADING_LIB "${CUDSS_THREADING_LIB}")
endif()
message(STATUS "CUDSS threading layer library: ${MFEM_CUDSS_THREADING_LIB}")
endif()
# Set the full name of the cuDSS communication library if MFEM use OpenMPI.
# The communication layer library (libcudss_commlayer_mpi.so) is located under the
# cuDSS library directory by default.
# The communication layer library is used pre-built communication layers for OpenMPI
# by default.
if (MFEM_USE_MPI)
find_file(
CUDSS_COMM_LIB
NAMES libcudss_commlayer_openmpi.so
PATHS ${CUDSS_LIBRARY_DIR}
NO_DEFAULT_PATH
)
if (NOT DEFINED MFEM_CUDSS_COMM_LIB AND CUDSS_COMM_LIB)
set(MFEM_CUDSS_COMM_LIB "${CUDSS_COMM_LIB}")
endif()
message(STATUS "CUDSS communication layer library: ${MFEM_CUDSS_COMM_LIB}")
endif()
+10 -8
View File
@@ -18,17 +18,19 @@
if (MFEM_FETCH_GSLIB OR MFEM_FETCH_TPLS)
enable_language(C)
set(GSLIB_FETCH_VERSION 1.0.9)
add_library(GSLIB STATIC IMPORTED)
# set options (technically flags because GSLIB does not use cmake)
string(TOUPPER "${CMAKE_BUILD_TYPE}" BUILD_TYPE)
set(GSLIB_FLAGS "${CMAKE_C_FLAGS} ${CMAKE_C_FLAGS_${BUILD_TYPE}}")
if (BUILD_SHARED_LIBS)
set(GSLIB_FLAGS "${GSLIB_FLAGS} -fPIC")
set(GSLIB_FETCH_VERSION 1.0.9)
set(GSLIB_C_FLAGS ${CMAKE_C_FLAGS_${BUILD_TYPE}})
if (CMAKE_C_FLAGS)
set(GSLIB_C_FLAGS "${CMAKE_C_FLAGS} ${CMAKE_C_FLAGS_${BUILD_TYPE}}")
endif()
if (BUILD_SHARED_LIBS)
set(GSLIB_C_FLAGS "${GSLIB_C_FLAGS} -fPIC")
endif()
add_library(GSLIB STATIC IMPORTED)
# define external project and create future include directory so it is present
# to pass CMake checks at end of MFEM configuration step
message(STATUS "Will fetch GSLIB ${GSLIB_FETCH_VERSION} to be built with ${GSLIB_FLAGS}")
message(STATUS "Will fetch GSLIB ${GSLIB_FETCH_VERSION} to be built with ${GSLIB_C_FLAGS}")
set(PREFIX ${CMAKE_BINARY_DIR}/fetch/gslib)
include(ExternalProject)
ExternalProject_Add(gslib
@@ -38,7 +40,7 @@ if (MFEM_FETCH_GSLIB OR MFEM_FETCH_TPLS)
UPDATE_DISCONNECTED TRUE
PREFIX ${PREFIX}
CONFIGURE_COMMAND ""
BUILD_COMMAND cd ${PREFIX}/src/gslib && $(MAKE) clean && $(MAKE) DESTDIR=${PREFIX} MPI=$<BOOL:${MFEM_USE_MPI}> "CFLAGS=${GSLIB_FLAGS}"
BUILD_COMMAND cd ${PREFIX}/src/gslib && $(MAKE) clean && $(MAKE) DESTDIR=${PREFIX} MPI=$<BOOL:${MFEM_USE_MPI}> "CFLAGS= ${GSLIB_C_FLAGS}"
INSTALL_COMMAND "")
file(MAKE_DIRECTORY ${PREFIX}/include)
# set imported library target properties
+1 -3
View File
@@ -44,9 +44,6 @@ if (MFEM_FETCH_HYPRE OR MFEM_FETCH_TPLS)
# set options and associated dependencies
set(HYPRE_CMAKE_OPTIONS "")
list(APPEND HYPRE_CMAKE_OPTIONS -DCMAKE_BUILD_TYPE:STRING=${CMAKE_BUILD_TYPE})
if (BUILD_SHARED_LIBS)
list(APPEND HYPRE_CMAKE_OPTIONS -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON)
endif()
# collect all HYPRE_ENABLE variables and pass them to hypre, assuming they are BOOL.
get_cmake_property(all_vars VARIABLES)
foreach(var ${all_vars})
@@ -98,6 +95,7 @@ if (MFEM_FETCH_HYPRE OR MFEM_FETCH_TPLS)
UPDATE_DISCONNECTED TRUE
SOURCE_SUBDIR src
PREFIX ${HYPRE_INSTALL}
BUILD_COMMAND ${CMAKE_COMMAND} --build . -- -j${CMAKE_BUILD_PARALLEL_LEVEL}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${HYPRE_INSTALL} -DCMAKE_INSTALL_LIBDIR:PATH=lib ${HYPRE_CMAKE_OPTIONS})
file(MAKE_DIRECTORY ${HYPRE_INSTALL}/include)
# set imported library target properties
+2 -10
View File
@@ -19,18 +19,10 @@
# - METIS_VERSION_5 (cache variable)
if (MFEM_FETCH_METIS OR MFEM_FETCH_TPLS)
enable_language(C)
set(METIS_FETCH_VERSION 4.0.3)
add_library(METIS STATIC IMPORTED)
# set options (technically flags because METIS does not use cmake)
set(METIS_FLAGS "-Wno-implicit-int -Wno-incompatible-pointer-types")
string(TOUPPER "${CMAKE_BUILD_TYPE}" BUILD_TYPE)
set(METIS_FLAGS "${METIS_FLAGS} ${CMAKE_C_FLAGS} ${CMAKE_C_FLAGS_${BUILD_TYPE}}")
if (BUILD_SHARED_LIBS)
set(METIS_FLAGS "${METIS_FLAGS} -fPIC")
endif()
# define external project
message(STATUS "Will fetch METIS ${METIS_FETCH_VERSION} to be built with ${METIS_FLAGS}")
message(STATUS "Will fetch METIS ${METIS_FETCH_VERSION} to be built with default options")
set(PREFIX ${CMAKE_BINARY_DIR}/fetch/metis)
include(ExternalProject)
ExternalProject_Add(metis
@@ -40,7 +32,7 @@ if (MFEM_FETCH_METIS OR MFEM_FETCH_TPLS)
UPDATE_DISCONNECTED TRUE
PREFIX ${PREFIX}
CONFIGURE_COMMAND tar -xzf ../metis/metis-${METIS_FETCH_VERSION}-mac.tgz --strip=1
BUILD_COMMAND $(MAKE) clean && $(MAKE) "OPTFLAGS=${METIS_FLAGS}"
BUILD_COMMAND $(MAKE) COPTIONS=-Wno-incompatible-pointer-types
INSTALL_COMMAND mkdir -p ${PREFIX}/lib && cp libmetis.a ${PREFIX}/lib/)
# set imported library target properties
add_dependencies(METIS metis)
+9 -9
View File
@@ -22,15 +22,15 @@ include(MfemCmakeUtilities)
mfem_find_package(SuiteSparse SuiteSparse SuiteSparse_DIR "" "" "" ""
"Paths to headers required by SuiteSparse."
"Libraries required by SuiteSparse."
ADD_COMPONENT "UMFPACK" "include;include/suitesparse;suitesparse" umfpack.h "lib" umfpack
ADD_COMPONENT "KLU" "include;include/suitesparse;suitesparse" klu.h "lib" klu
ADD_COMPONENT "AMD" "include;include/suitesparse;suitesparse" amd.h "lib" amd
ADD_COMPONENT "BTF" "include;include/suitesparse;suitesparse" btf.h "lib" btf
ADD_COMPONENT "CHOLMOD" "include;include/suitesparse;suitesparse" cholmod.h "lib" cholmod
ADD_COMPONENT "COLAMD" "include;include/suitesparse;suitesparse" colamd.h "lib" colamd
ADD_COMPONENT "CAMD" "include;include/suitesparse;suitesparse" camd.h "lib" camd
ADD_COMPONENT "CCOLAMD" "include;include/suitesparse;suitesparse" ccolamd.h "lib" ccolamd
ADD_COMPONENT "config" "include;include/suitesparse;suitesparse" SuiteSparse_config.h "lib"
ADD_COMPONENT "UMFPACK" "include;suitesparse" umfpack.h "lib" umfpack
ADD_COMPONENT "KLU" "include;suitesparse" klu.h "lib" klu
ADD_COMPONENT "AMD" "include;suitesparse" amd.h "lib" amd
ADD_COMPONENT "BTF" "include;suitesparse" btf.h "lib" btf
ADD_COMPONENT "CHOLMOD" "include;suitesparse" cholmod.h "lib" cholmod
ADD_COMPONENT "COLAMD" "include;suitesparse" colamd.h "lib" colamd
ADD_COMPONENT "CAMD" "include;suitesparse" camd.h "lib" camd
ADD_COMPONENT "CCOLAMD" "include;suitesparse" ccolamd.h "lib" ccolamd
ADD_COMPONENT "config" "include;suitesparse" SuiteSparse_config.h "lib"
suitesparseconfig)
if (SuiteSparse_FOUND AND METIS_VERSION_5)
+3 -3
View File
@@ -14,12 +14,12 @@
# - UMPIRE_LIBRARIES
# - UMPIRE_INCLUDE_DIRS
if (NOT umpire_ROOT AND UMPIRE_DIR)
set(umpire_ROOT ${UMPIRE_DIR})
if (NOT umpire_DIR AND UMPIRE_DIR)
set(umpire_DIR ${UMPIRE_DIR}/lib/cmake/umpire)
endif()
message(STATUS "Looking for UMPIRE ...")
message(STATUS " in UMPIRE_DIR = ${UMPIRE_DIR}")
message(STATUS " umpire_ROOT = ${umpire_ROOT}")
message(STATUS " umpire_DIR = ${umpire_DIR}")
find_package(umpire CONFIG)
set(UMPIRE_FOUND ${umpire_FOUND})
set(UMPIRE_LIBRARIES "umpire")
+17 -89
View File
@@ -701,6 +701,7 @@ endfunction(mfem_find_library)
# Extract compile and link options needed by the given target.
#
function(mfem_get_target_options Target CompileOptsVar LinkOptsVar)
if (NOT TARGET ${Target})
return()
endif()
@@ -798,12 +799,7 @@ function(mfem_get_target_options Target CompileOptsVar LinkOptsVar)
# message(STATUS "Lib = ${Lib}")
# Filter-out generator expressions
if (NOT ("${Lib}" MATCHES "^\\$"))
if(NOT ("${Lib}" STREQUAL "dl"))
list(APPEND LinkOpts "${Lib}")
else()
# for some reason libdl doesn't include the "-l"
list(APPEND LinkOpts "-ldl")
endif()
list(APPEND LinkOpts "${Lib}")
endif()
else()
mfem_get_target_options(${Lib} COpts LOpts)
@@ -892,18 +888,9 @@ function(mfem_export_mk_files)
set(${var} NO)
endif()
endforeach()
if (MFEM_USE_CUDA AND MFEM_EXPORT_GPU_CONFIG)
set(MFEM_CXX ${CMAKE_CUDA_COMPILER})
if(MFEM_CUDA_COMPILER_IS_NVCC)
set(MFEM_HOST_CXX ${CMAKE_CUDA_HOST_COMPILER})
else()
set(MFEM_HOST_CXX ${CMAKE_CXX_COMPILER})
endif()
else()
# mfem doesn't use enable_language(HIP)
set(MFEM_CXX ${CMAKE_CXX_COMPILER})
set(MFEM_HOST_CXX ${CMAKE_CXX_COMPILER})
endif()
# TODO: Add support for MFEM_USE_CUDA=YES
set(MFEM_CXX ${CMAKE_CXX_COMPILER})
set(MFEM_HOST_CXX ${MFEM_CXX})
set(MFEM_CPPFLAGS "")
get_target_property(cxx_std mfem CXX_STANDARD)
# For now, we ignore the setting of the CXX_EXTENSIONS property. If this
@@ -913,50 +900,6 @@ function(mfem_export_mk_files)
string(STRIP
"${cxx_std_flag} ${CMAKE_CXX_FLAGS_${BUILD_TYPE}} ${CMAKE_CXX_FLAGS}"
MFEM_CXXFLAGS)
if(MFEM_EXPORT_GPU_CONFIG)
if (MFEM_USE_CUDA)
set(MFEM_CXXFLAGS "${MFEM_CXXFLAGS} ${CMAKE_CUDA_FLAGS}")
if (MFEM_CUDA_COMPILER_IS_NVCC)
set(MFEM_CXXFLAGS "-x=cu ${MFEM_CXXFLAGS} -ccbin ${CMAKE_CXX_COMPILER} --forward-unknown-to-host-compiler")
# The following intentionally hides CUDA deprecation warnings
foreach(ENTRY IN LISTS CUDAToolkit_INCLUDE_DIRS)
set(MFEM_CXXFLAGS "${MFEM_CXXFLAGS} -isystem ${ENTRY}")
endforeach()
if (CMAKE_VERSION VERSION_GREATER_EQUAL 3.18.0)
# architecture flags not part of CMAKE_CUDA_FLAGS
if ("all" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}"
OR "native" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}"
OR "all-major" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}")
set(MFEM_CXXFLAGS "${MFEM_CXXFLAGS} -arch=${CMAKE_CUDA_ARCHITECTURES}")
else()
foreach (ENTRY IN LISTS CMAKE_CUDA_ARCHITECTURES)
set(MFEM_CXXFLAGS
"${MFEM_CXXFLAGS} -gencode arch=compute_${ENTRY},code=sm_${ENTRY}")
endforeach()
endif()
endif()
else()
set(MFEM_CXXFLAGS "${MFEM_CXXFLAGS} -xcuda --cuda-path=${CUDAToolkit_LIBRARY_ROOT}")
if (CMAKE_VERSION VERSION_GREATER_EQUAL 3.18.0)
# architecture flags not part of CMAKE_CUDA_FLAGS
if ("all" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}"
OR "native" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}"
OR "all-major" STREQUAL "${CMAKE_CUDA_ARCHITECTURES}")
# TODO: not supported
else()
foreach(ENTRY IN LISTS CMAKE_CUDA_ARCHITECTURES)
set(MFEM_CXXFLAGS "-cuda-gpu-arch=sm_${ENTRY} ${MFEM_CXXFLAGS}")
endforeach()
endif()
endif()
endif()
elseif (MFEM_USE_HIP)
set(MFEM_CXXFLAGS "${MFEM_CXXFLAGS} -xhip")
foreach(ENTRY IN LISTS CMAKE_HIP_ARCHITECTURES)
set(MFEM_CXXFLAGS "--offload-arch=${ENTRY} ${MFEM_CXXFLAGS}")
endforeach()
endif()
endif()
set(MFEM_TPLFLAGS "")
foreach(dir ${TPL_INCLUDE_DIRS})
set(MFEM_TPLFLAGS "${MFEM_TPLFLAGS} -I${dir}")
@@ -987,9 +930,6 @@ function(mfem_export_mk_files)
set(MFEM_SHARED NO)
set(MFEM_STATIC YES)
endif()
if (MFEM_USE_CUDA)
set(MFEM_EXT_LIBS "${MFEM_EXT_LIBS} -lcudart")
endif()
set(MFEM_BUILD_TAG "${CMAKE_SYSTEM}")
set(MFEM_PREFIX "${CMAKE_INSTALL_PREFIX}")
# For the next 4 variables, these are the values for the build-tree version of
@@ -998,15 +938,8 @@ function(mfem_export_mk_files)
set(MFEM_LIB_DIR "${PROJECT_BINARY_DIR}")
set(MFEM_TEST_MK "${PROJECT_SOURCE_DIR}/config/test.mk")
set(MFEM_CONFIG_EXTRA "MFEM_BUILD_DIR ?= ${PROJECT_BINARY_DIR}")
if (MFEM_USE_CUDA AND MFEM_EXPORT_GPU_CONFIG)
if (MFEM_CUDA_COMPILER_IS_NVCC)
set(MFEM_XLINKER "-Xlinker=")
else()
set(MFEM_XLINKER "${CMAKE_CUDA_LINKER_WRAPPER_FLAG}")
endif()
else()
set(MFEM_XLINKER "${CMAKE_CXX_LINKER_WRAPPER_FLAG}")
endif()
# TODO: CUDA/HIP support:
set(MFEM_XLINKER "${CMAKE_CXX_LINKER_WRAPPER_FLAG}")
set(MFEM_MPIEXEC ${MPIEXEC})
if (NOT MFEM_MPIEXEC)
set(MFEM_MPIEXEC "mpirun")
@@ -1054,21 +987,16 @@ function(mfem_export_mk_files)
# handle interfaces (e.g., SCOREC::apf)
if ("${lib}" MATCHES "SCOREC::.*" OR "${lib}" MATCHES "Ginkgo::.*" OR "${lib}" MATCHES "ParMoonolith::.*")
elseif (TARGET "${lib}")
mfem_get_target_options(${lib} CompileOpts2 LinkOpts2)
# remove generator expressions
string(GENEX_STRIP "${CompileOpts2}" CompileOpts)
string(GENEX_STRIP "${LinkOpts2}" LinkOpts)
mfem_get_target_options(${lib} CompileOpts LinkOpts)
# Removing duplicates may lead to issues:
# list(REMOVE_DUPLICATES CompileOpts)
# list(REMOVE_DUPLICATES LinkOpts)
# message(WARNING "${lib}[LinkOpts]: ${LinkOpts}")
# message(WARNING "${lib}[CompileOpts]: ${CompileOpts}")
foreach(LOpt IN LISTS LinkOpts)
set(MFEM_EXT_LIBS "${MFEM_EXT_LIBS} ${LOpt}")
endforeach()
foreach(COpt IN LISTS CompileOpts)
set(MFEM_TPLFLAGS "${MFEM_TPLFLAGS} ${COpt}")
endforeach()
string(REPLACE ";" " " COpts "${CompileOpts}")
string(REPLACE ";" " " LOpts "${LinkOpts}")
# message(STATUS "${lib}[COpts]: '${COpts}'")
# message(STATUS "${lib}[LOpts]: '${LOpts}'")
set(MFEM_TPLFLAGS "${MFEM_TPLFLAGS} ${COpts}")
set(MFEM_EXT_LIBS "${MFEM_EXT_LIBS} ${LOpts}")
# message(FATAL_ERROR "***** interface lib found ... exiting *****")
# handle static and shared libs
elseif ("${suffix}" STREQUAL "${CMAKE_SHARED_LIBRARY_SUFFIX}")
@@ -1076,7 +1004,7 @@ function(mfem_export_mk_files)
get_filename_component(fullLibName ${lib} NAME_WE)
string(REGEX REPLACE "^lib" "" libname ${fullLibName})
set(MFEM_EXT_LIBS
"${MFEM_EXT_LIBS} ${shared_link_flag}${dir} -L${dir} -l${libname}")
"${MFEM_EXT_LIBS} ${shared_link_flag}${dir} -L${dir} -l${libname}")
else()
set(MFEM_EXT_LIBS "${MFEM_EXT_LIBS} ${lib}")
endif()
@@ -1085,7 +1013,7 @@ function(mfem_export_mk_files)
# Create the build-tree version of 'config.mk'
configure_file(
"${PROJECT_SOURCE_DIR}/config/config.mk.in"
"${PROJECT_BINARY_DIR}/config/config.mk" @ONLY)
"${PROJECT_BINARY_DIR}/config/config.mk")
# Copy 'test.mk' from the source-tree to the build-tree
configure_file(
"${PROJECT_SOURCE_DIR}/config/test.mk"
@@ -1103,7 +1031,7 @@ function(mfem_export_mk_files)
# Create the install-tree version of 'config.mk'
configure_file(
"${PROJECT_SOURCE_DIR}/config/config.mk.in"
"${PROJECT_BINARY_DIR}/config/config-install.mk" @ONLY)
"${PROJECT_BINARY_DIR}/config/config-install.mk")
# Install rules for 'config.mk' and 'test.mk'
install(FILES ${PROJECT_SOURCE_DIR}/config/test.mk
-6
View File
@@ -157,10 +157,4 @@ constexpr real_t operator""_r(unsigned long long v)
#endif
#endif // MFEM_USE_MPI not defined
#ifndef MFEM_USE_CUDA
#ifdef MFEM_USE_CUDSS
#error Building with cuDSS (MFEM_USE_CUDSS=YES) requires CUDA (MFEM_USE_CUDA=YES)
#endif
#endif // MFEM_USE_CUDSS not defined
#endif // MFEM_CONFIG_HPP
-9
View File
@@ -108,15 +108,6 @@
// Enable MFEM functionality based on the STRUMPACK library.
// #define MFEM_USE_STRUMPACK
// Enable MFEM functionality based on the cuDSS library.
// #define MFEM_USE_CUDSS
// CUDSS communication layer library path
// #define MFEM_CUDSS_COMM_LIB "@MFEM_CUDSS_COMM_LIB@"
// CUDSS threading layer library path
// #define MFEM_CUDSS_THREADING_LIB "@MFEM_CUDSS_THREADING_LIB@"
// Enable MFEM features based on the Ginkgo library.
// #define MFEM_USE_GINKGO
-3
View File
@@ -36,9 +36,6 @@ MFEM_USE_SUPERLU = @MFEM_USE_SUPERLU@
MFEM_USE_SUPERLU5 = @MFEM_USE_SUPERLU5@
MFEM_USE_MUMPS = @MFEM_USE_MUMPS@
MFEM_USE_STRUMPACK = @MFEM_USE_STRUMPACK@
MFEM_USE_CUDSS = @MFEM_USE_CUDSS@
MFEM_CUDSS_COMM_LIB = @MFEM_CUDSS_COMM_LIB@
MFEM_CUDSS_THREADING_LIB = @MFEM_CUDSS_THREADING_LIB@
MFEM_USE_GINKGO = @MFEM_USE_GINKGO@
MFEM_USE_AMGX = @MFEM_USE_AMGX@
MFEM_USE_MAGMA = @MFEM_USE_MAGMA@
-1
View File
@@ -38,7 +38,6 @@ option(MFEM_USE_SUPERLU "Enable SuperLU_DIST usage" OFF)
option(MFEM_USE_SUPERLU5 "Use the old SuperLU_DIST 5.1 version" OFF)
option(MFEM_USE_MUMPS "Enable MUMPS usage" OFF)
option(MFEM_USE_STRUMPACK "Enable STRUMPACK usage" OFF)
option(MFEM_USE_CUDSS "Enable cuDSS usage" OFF)
option(MFEM_USE_GINKGO "Enable Ginkgo usage" OFF)
option(MFEM_USE_AMGX "Enable AmgX usage" OFF)
option(MFEM_USE_MAGMA "Enable MAGMA usage" OFF)
+1 -16
View File
@@ -18,7 +18,6 @@
# Some choices below are based on the OS type:
NOTMAC := $(subst Darwin,,$(shell uname -s))
ASTYLE_BIN = astyle
ETAGS_BIN = $(shell command -v etags 2> /dev/null)
EGREP_BIN = $(shell command -v egrep 2> /dev/null)
@@ -153,7 +152,6 @@ MFEM_USE_SUPERLU = NO
MFEM_USE_SUPERLU5 = NO
MFEM_USE_MUMPS = NO
MFEM_USE_STRUMPACK = NO
MFEM_USE_CUDSS = NO
MFEM_USE_GINKGO = NO
MFEM_USE_AMGX = NO
MFEM_USE_MAGMA = NO
@@ -369,19 +367,6 @@ STRUMPACK_OPT = -I$(STRUMPACK_DIR)/include $(SCOTCH_OPT)
STRUMPACK_LIB = -L$(STRUMPACK_DIR)/lib -lstrumpack $(MPI_FORTRAN_LIB)\
$(SCOTCH_LIB) $(SCALAPACK_LIB)
# CUDSS library configuration
CUDSS_DIR = @MFEM_DIR@/../cudss
CUDSS_INCLUDE_DIR = $(CUDSS_DIR)/include
CUDSS_LIBRARY_DIR = $(CUDSS_DIR)/lib
CUDSS_OPT = -I$(CUDSS_INCLUDE_DIR)
CUDSS_LIB = \
$(XLINKER)-rpath,$(CUDSS_LIBRARY_DIR) -L$(CUDSS_LIBRARY_DIR) -lcudss
# The cuDSS communication and threading libraries.
MFEM_CUDSS_COMM_LIB = $(abspath $(wildcard $(or $(CUDSS_COMM_LIB),\
$(subst @MFEM_DIR@,$(MFEM_DIR), $(CUDSS_LIBRARY_DIR)/libcudss_commlayer_openmpi.so))))
MFEM_CUDSS_THREADING_LIB = $(abspath $(wildcard $(or $(CUDSS_THREADING_LIB),\
$(subst @MFEM_DIR@,$(MFEM_DIR),$(CUDSS_LIBRARY_DIR)/libcudss_mtlayer_gomp.so))))
# Ginkgo library configuration
GINKGO_DIR = @MFEM_DIR@/../ginkgo/install
GINKGO_SEARCH_DIR = $(subst @MFEM_DIR@,$(MFEM_DIR),$(GINKGO_DIR))
@@ -635,7 +620,7 @@ PARELAG_LIB = -L$(PARELAG_DIR)/build/src -lParELAG
AXOM_DIR = @MFEM_DIR@/../axom
TRIBOL_DIR = @MFEM_DIR@/../tribol
TRIBOL_OPT = -I$(TRIBOL_DIR)/include -I$(AXOM_DIR)/include
TRIBOL_LIB = -L$(TRIBOL_DIR)/lib -ltribol -ltribol_shared -lredecomp -L$(AXOM_DIR)/lib -laxom_mint\
TRIBOL_LIB = -L$(TRIBOL_DIR)/lib -ltribol -lredecomp -L$(AXOM_DIR)/lib -laxom_mint\
-laxom_slam -laxom_slic -laxom_core
# Enzyme configuration
@@ -1,86 +0,0 @@
MFEM NURBS mesh v1.0
dimension
1
# Four segments with different NURBS orders, described via patches.
elements
4
1 1 0 1
2 1 2 3
3 1 4 5
4 1 6 7
boundary
0
edges
4
0 0 1
1 2 3
2 4 5
3 6 7
vertices
8
patches
# Patch 0: linear (order 1, 3 spans)
knotvectors
1
1 4 0 0 .4 .6 1 1
dimension
2
controlpoints
0.0 0.0 1.0
0.6 0.4 1.0
0.4 0.6 1.0
1.0 1.0 1.0
# Patch 1: quadratic (order 2, 2 spans)
knotvectors
1
2 4 0 0 0 .5 1 1 1
dimension
2
controlpoints
1.0 0.0 1.0
1.9 0.0 1.21
2.0 0.9 1.22
2.0 1.0 1.0
# Patch 2: cubic (order 3, 3 spans)
knotvectors
1
3 6 0 0 0 0 .33 .66 1 1 1 1
dimension
2
controlpoints
2.0 0.0 1.0
2.1 0.2 1.31
3.5 0.4 1.32
2.5 0.6 1.33
2.9 1.0 1.34
3.0 1.0 1.0
# Patch 3: quartic (order 4, 1 span)
knotvectors
1
4 5 0 0 0 0 0 1 1 1 1 1
dimension
2
controlpoints
3.0 0.0 1.0
3.45 0.5 1.41
3.50 1.0 1.42
3.75 0.8 1.43
4.0 0.0 1.0
-79
View File
@@ -1,79 +0,0 @@
MFEM NURBS mesh v1.0
#
# MFEM Geometry Types (see fem/geom.hpp):
#
# SEGMENT = 1
# SQUARE = 3
# CUBE = 5
#
dimension
1
# Three segments with different NURBS orders, described via patches.
elements
3
1 1 0 1
2 1 2 3
3 1 4 5
boundary
6
1 0 0
1 0 1
1 0 2
1 0 3
1 0 4
1 0 5
edges
3
0 0 1
1 2 3
2 4 5
vertices
6
patches
# Patch 0: linear (order 1, 2 control points)
knotvectors
1
1 2 0 0 1 1
dimension
2
controlpoints
0.0 0.0 1.0
1.0 1.0 1.0
# Patch 1: quadratic (order 2, 3 control points)
knotvectors
1
2 3 0 0 0 1 1 1
dimension
2
controlpoints
1.0 0.0 1.0
1.02 1.02 1.2
2.0 1.0 1.0
# Patch 2: cubic (order 3, 4 control points)
knotvectors
1
3 4 0 0 0 0 1 1 1 1
dimension
2
controlpoints
2.0 0.0 1.0
2.03 0.83 1.31
2.33 1.03 1.32
3.0 1.0 1.0
-72
View File
@@ -1,72 +0,0 @@
MFEM NURBS mesh v1.0
#
# MFEM Geometry Types (see fem/geom.hpp):
#
# SEGMENT = 1
# SQUARE = 3
# CUBE = 5
#
dimension
1
elements
3
1 1 0 1
2 1 2 3
3 1 4 5
boundary
6
1 0 0
1 0 1
1 0 2
1 0 3
1 0 4
1 0 5
edges
3
0 0 1
1 2 3
2 4 5
vertices
6
# Edge 0: linear (order 1, 2 control points)
# Edge 1: quadratic (order 2, 3 control points)
# Edge 2: cubic (order 3, 4 control points)
knotvectors
3
1 2 0 0 1 1
2 3 0 0 0 1 1 1
3 4 0 0 0 0 1 1 1 1
# One weight per control point, in the same order as the control points; (2 + 3 + 4) = 9 weights total
weights
1
1
1
1
1
1
1.2
1.31
1.32
FiniteElementSpace
FiniteElementCollection: NURBS
VDim: 2
Ordering: 1
0.0 0.0
1.0 1.0
1.0 0.0
2.0 1.0
2.0 0.0
3.0 1.0
1.02 1.02
2.03 0.83
2.33 1.03
-79
View File
@@ -1,79 +0,0 @@
MFEM NURBS mesh v1.0
#
# MFEM Geometry Types (see fem/geom.hpp):
#
# SEGMENT = 1
# SQUARE = 3
# CUBE = 5
#
dimension
1
# Three segments with different NURBS orders, described via patches.
elements
3
1 1 0 1
2 1 2 3
3 1 4 5
boundary
6
1 0 0
1 0 1
1 0 2
1 0 3
1 0 4
1 0 5
edges
3
0 0 1
1 2 3
2 4 5
vertices
6
patches
# Patch 0: linear (order 1, 2 control points)
knotvectors
1
1 2 0 0 1 1
dimension
3
controlpoints
0.0 0.0 0.01 1.0
1.0 1.0 1.01 1.0
# Patch 1: quadratic (order 2, 3 control points)
knotvectors
1
2 3 0 0 0 1 1 1
dimension
3
controlpoints
1.0 0.0 0.02 1.0
1.02 1.02 0.52 1.2
2.0 1.0 1.02 1.0
# Patch 2: cubic (order 3, 4 control points)
knotvectors
1
3 4 0 0 0 0 1 1 1 1
dimension
3
controlpoints
2.0 0.0 0.03 1.0
2.03 0.83 0.33 1.31
2.33 1.03 0.63 1.32
3.0 1.0 1.03 1.0
-72
View File
@@ -1,72 +0,0 @@
MFEM NURBS mesh v1.0
#
# MFEM Geometry Types (see fem/geom.hpp):
#
# SEGMENT = 1
# SQUARE = 3
# CUBE = 5
#
dimension
1
elements
3
1 1 0 1
2 1 2 3
3 1 4 5
boundary
6
1 0 0
1 0 1
1 0 2
1 0 3
1 0 4
1 0 5
edges
3
0 0 1
1 2 3
2 4 5
vertices
6
# Edge 0: linear (order 1, 2 control points)
# Edge 1: quadratic (order 2, 3 control points)
# Edge 2: cubic (order 3, 4 control points)
knotvectors
3
1 2 0 0 1 1
2 3 0 0 0 1 1 1
3 4 0 0 0 0 1 1 1 1
# One weight per control point, in the same order as the control points; (2 + 3 + 4) = 9 weights total
weights
1
1
1
1
1
1
1.2
1.31
1.32
FiniteElementSpace
FiniteElementCollection: NURBS
VDim: 3
Ordering: 1
0.0 0.0 0.01
1.0 1.0 1.01
1.0 0.0 0.02
2.0 1.0 1.02
2.0 0.0 0.03
3.0 1.0 1.03
1.02 1.02 0.52
2.03 0.83 0.33
2.33 1.03 0.63
-3
View File
@@ -190,8 +190,6 @@ namespace mfem {
* <a class="el" href="nurbs__ex1p_8cpp_source.html">1p</a>,
* <a class="el" href="nurbs__ex3_8cpp_source.html">3</a>,
* <a class="el" href="nurbs__ex5_8cpp_source.html">5</a>,
* <a class="el" href="nurbs__ex10_8cpp_source.html">10</a>,
* <a class="el" href="nurbs__ex10p_8cpp_source.html">10p</a>,
* <a class="el" href="nurbs__ex11p_8cpp_source.html">11p</a>, and
* <a class="el" href="nurbs__ex24_8cpp_source.html">24</a>,
* demonstrating howto perform NURBS-based Isogeometric Analysis.
@@ -200,7 +198,6 @@ namespace mfem {
* - <a class="el" href="nurbs__curveint_8cpp_source.html">NURBS Interpolation</a>: NURBS interpolation of given geometry
* - <a class="el" href="nurbs__naca__cmesh_8cpp_source.html">NURBS NACA Mesher</a>: generate NURBS based mesh around a NACA foil
* - <a class="el" href="nurbs__printfunc_8cpp_source.html">NURBS Printer</a>: print the NURBS-basis
* - <a class="el" href="nurbs__mesh_info_8cpp_source.html">NURBS Mesh info</a>: print the info of a NURBS mesh
*
* <H3>Miniapps</H3>
* - <a class="el" href="volta_8cpp_source.html">Volta</a>: simple electrostatics simulation code
+1 -1
View File
@@ -215,7 +215,7 @@ if (MFEM_ENABLE_TESTING)
add_test(NAME ex1p_ceed_np=${MFEM_MPI_NP}
COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
${MPIEXEC_PREFLAGS}
$<TARGET_FILE:ex1p> "-no-vis" "-d" "ceed-cpu" "-pa" "-a"
$<TARGET_FILE:ex1p> "-no-vis" "-d ceed-cpu" "-pa" "-a"
${MPIEXEC_POSTFLAGS})
endif()
endif()
+1 -1
View File
@@ -64,7 +64,7 @@ PARALLEL_NAME := Parallel AMGX example
$(MFEM_LIB_FILE):
$(error The MFEM library is not build)
clean: clean-build clean-exec
clean: clean-build
clean-build:
rm -f *.o *~ $(SEQ_EXAMPLES) $(PAR_EXAMPLES)
+3 -3
View File
@@ -64,12 +64,12 @@ ex1p-test-par: ex1p
$(MFEM_LIB_FILE):
$(error The MFEM library is not built)
clean: clean-build clean-exec
clean: clean-build clean-exec $(SUBDIRS_CLEAN)
clean-build:
rm -f *.o *~ $(SEQ_EXAMPLES) $(PAR_EXAMPLES)
rm -rf *.dSYM *.TVD.*breakpoints
clean-exec:
@rm -f refined.mesh mesh.*
@rm -f sol.*
@rm -f refined.mesh displaced.mesh mesh.* ex5.mesh
@rm -f sphere_refined.* sol.* sol_u.* sol_p.* sol_r.* sol_i.*
+21 -34
View File
@@ -50,10 +50,6 @@
// ex1 -m ../data/beam-tet.mesh -pa -d ceed-cpu
// ex1 -m ../data/beam-tet.mesh -pa -d ceed-cuda:/gpu/cuda/ref
//
// Device simplices sample runs:
// ex1 -pa -d gpu -m ../data/inline-tet.mesh
// ex1 -pa -d gpu -m ../data/inline-tri.mesh
//
// Description: This example code demonstrates the use of MFEM to define a
// simple finite element discretization of the Poisson problem
// -Delta u = 1 with homogeneous Dirichlet boundary conditions.
@@ -142,25 +138,25 @@ int main(int argc, char *argv[])
}
// 5. Define a finite element space on the mesh. Here we use continuous
// Lagrange finite elements of the specified order.
// - If order < 1, we instead use an isoparametric/isogeometric space.
// - If the mesh is simplicial and partial assembly is requested,
// we use the positive basis, which supports device execution.
// Lagrange finite elements of the specified order. If order < 1, we
// instead use an isoparametric/isogeometric space.
FiniteElementCollection *fec;
auto basis_type = (pa && mesh.IsSimplexMesh()) ?
BasisType::Positive : BasisType::GaussLobatto;
bool delete_fec;
if (order > 0)
{
fec = new H1_FECollection(order, dim, basis_type);
fec = new H1_FECollection(order, dim);
delete_fec = true;
}
else if (mesh.GetNodes())
{
fec = mesh.GetNodes()->OwnFEC();
delete_fec = false;
cout << "Using isoparametric FEs: " << fec->Name() << endl;
}
else
{
fec = new H1_FECollection(order = 1, dim, basis_type);
fec = new H1_FECollection(order = 1, dim);
delete_fec = true;
}
FiniteElementSpace fespace(&mesh, fec);
cout << "Number of finite element unknowns: "
@@ -228,29 +224,17 @@ int main(int argc, char *argv[])
// 11. Solve the linear system A X = B.
if (!pa)
{
#ifdef MFEM_USE_CUDSS
if (Device::Allows(Backend::CUDA_MASK))
{
// Use cuDSS to solve the system.
CuDSSSolver cudss_solver;
cudss_solver.SetOperator(*A);
cudss_solver.Mult(B, X);
}
else
#endif
{
#ifndef MFEM_USE_SUITESPARSE
// Use a simple symmetric Gauss-Seidel preconditioner with PCG.
GSSmoother M((SparseMatrix&)(*A));
PCG(*A, M, B, X, 1, 200, 1e-12, 0.0);
// Use a simple symmetric Gauss-Seidel preconditioner with PCG.
GSSmoother M((SparseMatrix&)(*A));
PCG(*A, M, B, X, 1, 200, 1e-12, 0.0);
#else
// If MFEM was compiled with SuiteSparse, use UMFPACK to solve the system.
UMFPackSolver umf_solver;
umf_solver.Control[UMFPACK_ORDERING] = UMFPACK_ORDERING_METIS;
umf_solver.SetOperator(*A);
umf_solver.Mult(B, X);
// If MFEM was compiled with SuiteSparse, use UMFPACK to solve the system.
UMFPackSolver umf_solver;
umf_solver.Control[UMFPACK_ORDERING] = UMFPACK_ORDERING_METIS;
umf_solver.SetOperator(*A);
umf_solver.Mult(B, X);
#endif
}
}
else
{
@@ -289,14 +273,17 @@ int main(int argc, char *argv[])
if (visualization)
{
char vishost[] = "localhost";
int visport = 19916;
int visport = 19916;
socketstream sol_sock(vishost, visport);
sol_sock.precision(8);
sol_sock << "solution\n" << mesh << x << flush;
}
// 15. Free the used memory.
if (order > 0) { delete fec; }
if (delete_fec)
{
delete fec;
}
return 0;
}
+2 -2
View File
@@ -5,9 +5,9 @@
// Sample runs:
// mpirun -np 4 ex12p -m ../data/beam-tri.mesh
// mpirun -np 4 ex12p -m ../data/beam-quad.mesh
// mpirun -np 4 ex12p -m ../data/beam-tet.mesh -s 464 -n 10 -o 2 -elast
// mpirun -np 4 ex12p -m ../data/beam-tet.mesh -s 462 -n 10 -o 2 -elast
// mpirun -np 4 ex12p -m ../data/beam-hex.mesh -s 3878
// mpirun -np 4 ex12p -m ../data/beam-wedge.mesh -s 82
// mpirun -np 4 ex12p -m ../data/beam-wedge.mesh -s 81
// mpirun -np 4 ex12p -m ../data/beam-tri.mesh -s 3877 -o 2 -sys
// mpirun -np 4 ex12p -m ../data/beam-quad.mesh -s 4544 -n 6 -o 3 -elast
// mpirun -np 4 ex12p -m ../data/beam-quad-nurbs.mesh
+34 -60
View File
@@ -42,11 +42,7 @@
// mpirun -np 4 ex1p -pa -d ceed-cuda:/gpu/cuda/shared
// mpirun -np 4 ex1p -pa -d ceed-cuda:/gpu/cuda/shared -m ../data/square-mixed.mesh
// mpirun -np 4 ex1p -pa -d ceed-cuda:/gpu/cuda/shared -m ../data/fichera-mixed.mesh
// mpirun -np 4 ex1p -pa -d ceed-cpu -m ../data/beam-tet.mesh
//
// Device simplices sample runs:
// mpirun -np 4 ex1p -pa -d gpu -m ../data/inline-tet.mesh
// mpirun -np 4 ex1p -pa -d gpu -m ../data/inline-tri.mesh
// mpirun -np 4 ex1p -m ../data/beam-tet.mesh -pa -d ceed-cpu
//
// Description: This example code demonstrates the use of MFEM to define a
// simple finite element discretization of the Poisson problem
@@ -87,9 +83,6 @@ int main(int argc, char *argv[])
const char *device_config = "cpu";
bool visualization = true;
bool algebraic_ceed = false;
#ifdef MFEM_USE_CUDSS
bool cudss_solver = false;
#endif
OptionsParser args(argc, argv);
args.AddOption(&mesh_file, "-m", "--mesh",
@@ -109,10 +102,6 @@ int main(int argc, char *argv[])
args.AddOption(&algebraic_ceed, "-a", "--algebraic",
"-no-a", "--no-algebraic",
"Use algebraic Ceed solver");
#endif
#ifdef MFEM_USE_CUDSS
args.AddOption(&cudss_solver, "-cudss", "--cudss-solver", "-no-cudss",
"--no-cudss-solver", "Use the cuDSS Solver.");
#endif
args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
"--no-visualization",
@@ -169,20 +158,19 @@ int main(int argc, char *argv[])
}
// 7. Define a parallel finite element space on the parallel mesh. Here we
// use continuous Lagrange finite elements of the specified order.
// - If order < 1, we instead use an isoparametric/isogeometric space.
// - If the mesh is simplicial and partial assembly is requested,
// we use the positive basis, which supports device execution.
// use continuous Lagrange finite elements of the specified order. If
// order < 1, we instead use an isoparametric/isogeometric space.
FiniteElementCollection *fec;
auto basis_type = (pa && pmesh.IsSimplexMesh()) ?
BasisType::Positive : BasisType::GaussLobatto;
bool delete_fec;
if (order > 0)
{
fec = new H1_FECollection(order, dim, basis_type);
fec = new H1_FECollection(order, dim);
delete_fec = true;
}
else if (pmesh.GetNodes())
{
fec = pmesh.GetNodes()->OwnFEC();
delete_fec = false;
if (myid == 0)
{
cout << "Using isoparametric FEs: " << fec->Name() << endl;
@@ -190,7 +178,8 @@ int main(int argc, char *argv[])
}
else
{
fec = new H1_FECollection(order = 1, dim, basis_type);
fec = new H1_FECollection(order = 1, dim);
delete_fec = true;
}
ParFiniteElementSpace fespace(&pmesh, fec);
HYPRE_BigInt size = fespace.GlobalTrueVSize();
@@ -259,51 +248,33 @@ int main(int argc, char *argv[])
// 13. Solve the linear system A X = B.
// * With full assembly, use the BoomerAMG preconditioner from hypre.
// * With partial assembly, use Jacobi smoothing, for now.
#ifdef MFEM_USE_CUDSS
if (!pa && (Device::Allows(Backend::CUDA_MASK) && cudss_solver))
Solver *prec = NULL;
if (pa)
{
// Solve using a direct solver with cuDSS
CuDSSSolver cudss_solver(MPI_COMM_WORLD);
cudss_solver.SetMatrixSymType(
CuDSSSolver::SYMMETRIC_POSITIVE_DEFINITE);
cudss_solver.SetMatrixViewType(CuDSSSolver::UPPER);
cudss_solver.SetOperator(*A);
cudss_solver.Mult(B, X);
}
else
#endif
{
Solver *prec = NULL;
if (pa)
if (UsesTensorBasis(fespace))
{
if (UsesTensorBasis(fespace))
if (algebraic_ceed)
{
if (algebraic_ceed)
{
prec = new ceed::AlgebraicSolver(a, ess_tdof_list);
}
else
{
prec = new OperatorJacobiSmoother(a, ess_tdof_list);
}
prec = new ceed::AlgebraicSolver(a, ess_tdof_list);
}
else
{
prec = new OperatorJacobiSmoother(a, ess_tdof_list);
}
}
else
{
prec = new HypreBoomerAMG;
}
CGSolver cg(MPI_COMM_WORLD);
cg.SetRelTol(1e-12);
cg.SetMaxIter(2000);
cg.SetPrintLevel(1);
if (prec)
{
cg.SetPreconditioner(*prec);
}
cg.SetOperator(*A);
cg.Mult(B, X);
delete prec;
}
else
{
prec = new HypreBoomerAMG;
}
CGSolver cg(MPI_COMM_WORLD);
cg.SetRelTol(1e-12);
cg.SetMaxIter(2000);
cg.SetPrintLevel(1);
if (prec) { cg.SetPreconditioner(*prec); }
cg.SetOperator(*A);
cg.Mult(B, X);
delete prec;
// 14. Recover the parallel grid function corresponding to X. This is the
// local finite element solution on each processor.
@@ -337,7 +308,10 @@ int main(int argc, char *argv[])
}
// 17. Free the used memory.
if (order > 0) { delete fec; }
if (delete_fec)
{
delete fec;
}
return 0;
}
+1 -1
View File
@@ -119,7 +119,7 @@ int main(int argc, char *argv[])
}
LinearForm b(&fespace);
b.AddBoundaryIntegrator(new VectorBoundaryLFIntegrator(f));
b.AddDomainIntegrator(new VectorBoundaryLFIntegrator(f));
// 6. Set up the bilinear form a(.,.) on the finite element space
// corresponding to the linear elasticity integrator with piece-wise
+1 -1
View File
@@ -140,7 +140,7 @@ int main(int argc, char *argv[])
}
ParLinearForm b(&fespace);
b.AddBoundaryIntegrator(new VectorBoundaryLFIntegrator(f));
b.AddDomainIntegrator(new VectorBoundaryLFIntegrator(f));
// 6. Set up the bilinear form a(.,.) on the finite element space
// corresponding to the linear elasticity integrator with piece-wise
+9 -27
View File
@@ -302,21 +302,15 @@ int main(int argc, char *argv[])
char vishost[] = "localhost";
int visport = 19916;
socketstream sol_sock_r(vishost, visport);
socketstream sol_sock_i(vishost, visport);
sol_sock_r << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_r.precision(8);
sol_sock_i.precision(8);
sol_sock_r << "solution\n" << *pmesh << u_exact->real()
<< "window_title 'Exact: Real Part'" << flush;
// Make sure all ranks have sent their real solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
socketstream sol_sock_i(vishost, visport);
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i.precision(8);
sol_sock_i << "solution\n" << *pmesh << u_exact->imag()
<< "window_title 'Exact: Imaginary Part'" << flush;
// Make sure all ranks have sent their imaginary solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
}
// 11. Set up the parallel sesquilinear form a(.,.) on the finite element
@@ -540,21 +534,15 @@ int main(int argc, char *argv[])
char vishost[] = "localhost";
int visport = 19916;
socketstream sol_sock_r(vishost, visport);
socketstream sol_sock_i(vishost, visport);
sol_sock_r << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_r.precision(8);
sol_sock_i.precision(8);
sol_sock_r << "solution\n" << *pmesh << u.real()
<< "window_title 'Solution: Real Part'" << flush;
// Make sure all ranks have sent their real solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
socketstream sol_sock_i(vishost, visport);
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i.precision(8);
sol_sock_i << "solution\n" << *pmesh << u.imag()
<< "window_title 'Solution: Imaginary Part'" << flush;
// Make sure all ranks have sent their imaginary solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
}
if (visualization && exact_sol)
{
@@ -563,21 +551,15 @@ int main(int argc, char *argv[])
char vishost[] = "localhost";
int visport = 19916;
socketstream sol_sock_r(vishost, visport);
socketstream sol_sock_i(vishost, visport);
sol_sock_r << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_r.precision(8);
sol_sock_i.precision(8);
sol_sock_r << "solution\n" << *pmesh << u_exact->real()
<< "window_title 'Error: Real Part'" << flush;
// Make sure all ranks have sent their real solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
socketstream sol_sock_i(vishost, visport);
sol_sock_i << "parallel " << num_procs << " " << myid << "\n";
sol_sock_i.precision(8);
sol_sock_i << "solution\n" << *pmesh << u_exact->imag()
<< "window_title 'Error: Imaginary Part'" << flush;
// Make sure all ranks have sent their imaginary solution before initiating
// another set of GLVis connections (one from each rank):
MPI_Barrier(pmesh->GetComm());
}
if (visualization)
{
-9
View File
@@ -95,15 +95,6 @@ int main(int argc, char *argv[])
args.PrintOptions(cout);
}
if (amg_elast && !static_cond && reorder_space)
{
if (myid == 0)
cerr << "\nThe AMG elasticity solver requires ordering byVDIM! "
<< "Ignoring the specified option -nodes/--by-nodes.\n"
<< endl;
reorder_space = false;
}
// 3. Enable hardware devices such as GPUs, and programming models such as
// CUDA, OCCA, RAJA and OpenMP based on command line options.
Device device(device_config);
+52 -11
View File
@@ -5,8 +5,8 @@
// Sample runs:
// ex37 -alpha 10
// ex37 -alpha 10 -pv
// ex37 -lambda 0.1 -mu 0.1 -growth 1
// ex37 -o 2 -alpha 10.0 -mi 50 -vf 0.4 -ntol 1e-5 -growth 1.5
// ex37 -lambda 0.1 -mu 0.1
// ex37 -o 2 -alpha 5.0 -mi 50 -vf 0.4 -ntol 1e-5
// ex37 -r 6 -o 1 -alpha 25.0 -epsilon 0.02 -mi 50 -ntol 1e-5
//
// Description: This example code demonstrates the use of MFEM to solve a
@@ -55,6 +55,53 @@
using namespace std;
using namespace mfem;
/**
* @brief Bregman projection of ρ = sigmoid(ψ) onto the subspace
* ∫_Ω ρ dx = θ vol(Ω) as follows:
*
* 1. Compute the root of the R → R function
* f(c) = ∫_Ω sigmoid(ψ + c) dx - θ vol(Ω)
* 2. Set ψ ← ψ + c.
*
* @param psi a GridFunction to be updated
* @param target_volume θ vol(Ω)
* @param tol Newton iteration tolerance
* @param max_its Newton maximum iteration number
* @return real_t Final volume, ∫_Ω sigmoid(ψ)
*/
real_t proj(GridFunction &psi, real_t target_volume, real_t tol=1e-12,
int max_its=10)
{
MappedGridFunctionCoefficient sigmoid_psi(&psi, sigmoid);
MappedGridFunctionCoefficient der_sigmoid_psi(&psi, der_sigmoid);
LinearForm int_sigmoid_psi(psi.FESpace());
int_sigmoid_psi.AddDomainIntegrator(new DomainLFIntegrator(sigmoid_psi));
LinearForm int_der_sigmoid_psi(psi.FESpace());
int_der_sigmoid_psi.AddDomainIntegrator(new DomainLFIntegrator(
der_sigmoid_psi));
bool done = false;
for (int k=0; k<max_its; k++) // Newton iteration
{
int_sigmoid_psi.Assemble(); // Recompute f(c) with updated ψ
const real_t f = int_sigmoid_psi.Sum() - target_volume;
int_der_sigmoid_psi.Assemble(); // Recompute df(c) with updated ψ
const real_t df = int_der_sigmoid_psi.Sum();
const real_t dc = -f/df;
psi += dc;
if (abs(dc) < tol) { done = true; break; }
}
if (!done)
{
mfem_warning("Projection reached maximum iteration without converging. "
"Result may not be accurate.");
}
int_sigmoid_psi.Assemble();
return int_sigmoid_psi.Sum();
}
/*
* ---------------------------------------------------------------
* ALGORITHM PREAMBLE
@@ -133,11 +180,10 @@ int main(int argc, char *argv[])
int ref_levels = 5;
int order = 2;
real_t alpha = 1.0;
real_t growth = 2;
real_t epsilon = 0.01;
real_t vol_fraction = 0.5;
int max_it = 1e3;
real_t itol = 1e-2;
real_t itol = 1e-1;
real_t ntol = 1e-4;
real_t rho_min = 1e-6;
real_t lambda = 1.0;
@@ -152,8 +198,6 @@ int main(int argc, char *argv[])
"Order (degree) of the finite elements.");
args.AddOption(&alpha, "-alpha", "--alpha-step-length",
"Step length for gradient descent.");
args.AddOption(&growth, "-growth", "--alpha-growth-rate",
"Growth rate of step length for gradient descent.");
args.AddOption(&epsilon, "-epsilon", "--epsilon-thickness",
"Length scale for ρ.");
args.AddOption(&max_it, "-mi", "--max-it",
@@ -288,7 +332,6 @@ int main(int argc, char *argv[])
}
FilterSolver->SetEssentialBoundary(ess_bdr_filter);
FilterSolver->SetupFEM();
FilterSolver->AssembleDiffusionBilinear();
BilinearForm mass(&control_fes);
mass.AddDomainIntegrator(new InverseIntegrator(new MassIntegrator(one)));
@@ -342,7 +385,7 @@ int main(int argc, char *argv[])
// 11. Iterate:
for (int k = 1; k <= max_it; k++)
{
if (k > 1) { alpha = std::pow((real_t) k,growth); }
if (k > 1) { alpha *= ((real_t) k) / ((real_t) k-1); }
mfem::out << "\nStep = " << k << std::endl;
@@ -379,9 +422,7 @@ int main(int argc, char *argv[])
// Step 5 - Update design variable ψ ← proj(ψ - αG)
psi.Add(-alpha, grad);
GridFunction alpha_grad(grad);
alpha_grad *= alpha;
const real_t material_volume = proj(psi, alpha_grad, target_volume);
const real_t material_volume = proj(psi, target_volume);
// Compute ||ρ - ρ_old|| in control fes.
real_t norm_increment = zerogf.ComputeL1Error(succ_diff_rho);
+23 -183
View File
@@ -137,7 +137,7 @@ public:
exponent(exponent_), rho_min(rho_min_)
{
MFEM_ASSERT(rho_min_ >= 0.0, "rho_min must be >= 0");
MFEM_ASSERT(rho_min_ < 1.0, "rho_min must be < 1");
MFEM_ASSERT(rho_min_ < 1.0, "rho_min must be > 1");
MFEM_ASSERT(u, "displacement field is not set");
MFEM_ASSERT(rho_filter, "density field is not set");
}
@@ -231,12 +231,9 @@ private:
FiniteElementCollection * fec = nullptr;
FiniteElementSpace * fes = nullptr;
Array<int> ess_bdr;
Array<int> ess_tdof_list;
Array<int> neumann_bdr;
GridFunction * u = nullptr;
LinearForm * b = nullptr;
BilinearForm * a = nullptr;
OperatorPtr A;
bool parallel;
#ifdef MFEM_USE_MPI
ParMesh * pmesh = nullptr;
@@ -270,8 +267,6 @@ public:
void ResetFEM();
void SetupFEM();
void UpdateEssentialTDofs();
void AssembleDiffusionBilinear(bool update_ess_tdofs=true);
void Solve();
GridFunction * GetFEMSolution();
LinearForm * GetLinearForm() {return b;}
@@ -376,130 +371,6 @@ public:
};
/**
* @brief Bregman projection of ρ = sigmoid(ψ) onto the subspace
* ∫_Ω ρ dx = θ vol(Ω) as follows:
*
* 1. Compute the root of the R → R function
* f(c) = ∫_Ω sigmoid(ψ + c) dx - θ vol(Ω)
* using the Illinois method
* 2. Set ψ ← ψ + c.
*
* @param psi a GridFunction to be updated
* @param alpha_grad alpha multiplied by gradient
* @param target_volume θ vol(Ω)
* @param tol Illinois iteration tolerance
* @param max_its Illinois maximum iteration number
* @return real_t Final volume (∫_Ω sigmoid(ψ) dx)
*/
real_t proj(GridFunction &psi, GridFunction &alpha_grad, real_t target_volume,
real_t tol = 1e-12, int max_its = 100)
{
#ifdef MFEM_USE_MPI
FiniteElementSpace *fes = psi.FESpace();
ParFiniteElementSpace *pfes = dynamic_cast<ParFiniteElementSpace*>(fes);
#endif
ConstantCoefficient zero_cf(0.0);
real_t a = -alpha_grad.ComputeMaxError(zero_cf);
real_t b = -a;
real_t y = 0.0;
MappedGridFunctionCoefficient sigmoid_psi(
&psi, [&y](const real_t x) { return sigmoid(x + y); });
std::unique_ptr<LinearForm> int_sigmoid_psi;
#ifdef MFEM_USE_MPI
ParGridFunction *par_psi = dynamic_cast<ParGridFunction *>(&psi);
if (par_psi)
{
int_sigmoid_psi.reset(new ParLinearForm(par_psi->ParFESpace()));
}
else
{
int_sigmoid_psi.reset(new LinearForm(psi.FESpace()));
}
#else
int_sigmoid_psi.reset(new LinearForm(psi.FESpace()));
#endif
int_sigmoid_psi->AddDomainIntegrator(new DomainLFIntegrator(sigmoid_psi));
y = a;
int_sigmoid_psi->Assemble();
real_t f_a = int_sigmoid_psi->Sum(); // f_a := f(a) + θ vol(Ω)
y = b;
int_sigmoid_psi->Assemble();
real_t f_b = int_sigmoid_psi->Sum(); // f_b := f(b) + θ vol(Ω)
#ifdef MFEM_USE_MPI
if (pfes)
{
MPI_Allreduce(MPI_IN_PLACE, &f_a, 1, MPITypeMap<real_t>::mpi_type,
MPI_SUM, MPI_COMM_WORLD);
MPI_Allreduce(MPI_IN_PLACE, &f_b, 1, MPITypeMap<real_t>::mpi_type,
MPI_SUM, MPI_COMM_WORLD);
}
#endif
f_a -= target_volume; // f_a := f(a)
f_b -= target_volume; // f_b := f(b)
real_t c = 0.0;
real_t f_c = 0.0;
int side = 0;
bool done = false;
for (int k=0; k < max_its; k++)
{
c = (f_a * b - f_b * a) / (f_a - f_b);
if (abs(b - a) < tol * abs(b + a)) { done = true; break; }
y = c;
int_sigmoid_psi->Assemble();
f_c = int_sigmoid_psi->Sum(); // f_c := f(c) + θ vol(Ω)
#ifdef MFEM_USE_MPI
if (pfes)
{
MPI_Allreduce(MPI_IN_PLACE, &f_c, 1, MPITypeMap<real_t>::mpi_type,
MPI_SUM, MPI_COMM_WORLD);
}
#endif
f_c -= target_volume; // f_c := f(c)
if (f_c * f_b > 0)
{
b = c;
f_b = f_c;
if (side == -1) { f_a /= 2.0; }
side = -1;
}
else if (f_c * f_a > 0)
{
a = c;
f_a = f_c;
if (side == 1) { f_b /= 2.0; }
side = 1;
}
else
{
done = true; break;
}
}
if (!done)
{
mfem_warning("Projection reached maximum iteration without converging. "
"Result may not be accurate.");
}
y = 0.0;
psi += c;
int_sigmoid_psi->Assemble();
real_t material_volume = int_sigmoid_psi->Sum();
#ifdef MFEM_USE_MPI
if (pfes)
{
MPI_Allreduce(MPI_IN_PLACE, &material_volume, 1,
MPITypeMap<real_t>::mpi_type, MPI_SUM, MPI_COMM_WORLD);
}
#endif
return material_volume;
}
// Poisson solver
@@ -551,8 +422,12 @@ void DiffusionSolver::SetupFEM()
}
}
void DiffusionSolver::UpdateEssentialTDofs()
void DiffusionSolver::Solve()
{
OperatorPtr A;
Vector B, X;
Array<int> ess_tdof_list;
#ifdef MFEM_USE_MPI
if (parallel)
{
@@ -565,39 +440,7 @@ void DiffusionSolver::UpdateEssentialTDofs()
#else
fes->GetEssentialTrueDofs(ess_bdr,ess_tdof_list);
#endif
}
void DiffusionSolver::AssembleDiffusionBilinear(bool update_ess_tdofs)
{
if (update_ess_tdofs)
{
UpdateEssentialTDofs();
}
#ifdef MFEM_USE_MPI
if (parallel)
{
a = new ParBilinearForm(pfes);
}
else
{
a = new BilinearForm(fes);
}
#else
a = new BilinearForm(fes);
#endif
a->AddDomainIntegrator(new DiffusionIntegrator(*diffcf));
if (masscf)
{
a->AddDomainIntegrator(new MassIntegrator(*masscf));
}
a->Assemble();
a->FormSystemMatrix(ess_tdof_list, A);
}
void DiffusionSolver::Solve()
{
Vector B, X;
*u=0.0;
if (b)
{
delete b;
@@ -632,33 +475,31 @@ void DiffusionSolver::Solve()
b->Assemble();
*u=0.0;
if (essbdr_cf)
{
u->ProjectBdrCoefficient(*essbdr_cf,ess_bdr);
}
BilinearForm * a = nullptr;
#ifdef MFEM_USE_MPI
if (parallel)
{
X.SetSize(pfes->TrueVSize());
B.SetSize(pfes->TrueVSize());
dynamic_cast<ParGridFunction*>(u)->ParallelAssemble(X);
dynamic_cast<ParLinearForm*>(b)->ParallelAssemble(B);
dynamic_cast<ParBilinearForm*>(a)->ParallelEliminateTDofsInRHS(
ess_tdof_list, X, B);
a = new ParBilinearForm(pfes);
}
else
{
X.NewDataAndSize(u->GetData(), u->Size());
B.NewDataAndSize(b->GetData(), b->Size());
a->EliminateVDofsInRHS(ess_tdof_list, X, B);
a = new BilinearForm(fes);
}
#else
X.NewDataAndSize(u->GetData(), u->Size());
B.NewDataAndSize(b->GetData(), b->Size());
a->EliminateVDofsInRHS(ess_tdof_list, X, B);
a = new BilinearForm(fes);
#endif
a->AddDomainIntegrator(new DiffusionIntegrator(*diffcf));
if (masscf)
{
a->AddDomainIntegrator(new MassIntegrator(*masscf));
}
a->Assemble();
if (essbdr_cf)
{
u->ProjectBdrCoefficient(*essbdr_cf,ess_bdr);
}
a->FormLinearSystem(ess_tdof_list, *u, *b, A, X, B);
CGSolver * cg = nullptr;
Solver * M = nullptr;
@@ -687,6 +528,7 @@ void DiffusionSolver::Solve()
delete M;
delete cg;
a->RecoverFEMSolution(X, *b, *u);
delete a;
}
GridFunction * DiffusionSolver::GetFEMSolution()
@@ -718,8 +560,6 @@ DiffusionSolver::~DiffusionSolver()
#endif
delete fec; fec = nullptr;
delete b;
A.Clear();
delete a;
}
+60 -11
View File
@@ -4,8 +4,8 @@
//
// Sample runs:
// mpirun -np 4 ex37p -alpha 10 -pv
// mpirun -np 4 ex37p -lambda 0.1 -mu 0.1 -growth 1
// mpirun -np 4 ex37p -o 2 -alpha 10.0 -mi 50 -vf 0.4 -ntol 1e-5 -growth 1.5
// mpirun -np 4 ex37p -lambda 0.1 -mu 0.1
// mpirun -np 4 ex37p -o 2 -alpha 5.0 -mi 50 -vf 0.4 -ntol 1e-5
// mpirun -np 4 ex37p -r 6 -o 2 -alpha 10.0 -epsilon 0.02 -mi 50 -ntol 1e-5
//
// Description: This example code demonstrates the use of MFEM to solve a
@@ -54,6 +54,61 @@
using namespace std;
using namespace mfem;
/**
* @brief Bregman projection of ρ = sigmoid(ψ) onto the subspace
* ∫_Ω ρ dx = θ vol(Ω) as follows:
*
* 1. Compute the root of the R → R function
* f(c) = ∫_Ω sigmoid(ψ + c) dx - θ vol(Ω)
* 2. Set ψ ← ψ + c.
*
* @param psi a GridFunction to be updated
* @param target_volume θ vol(Ω)
* @param tol Newton iteration tolerance
* @param max_its Newton maximum iteration number
* @return real_t Final volume, ∫_Ω sigmoid(ψ)
*/
real_t proj(ParGridFunction &psi, real_t target_volume, real_t tol=1e-12,
int max_its=10)
{
MappedGridFunctionCoefficient sigmoid_psi(&psi, sigmoid);
MappedGridFunctionCoefficient der_sigmoid_psi(&psi, der_sigmoid);
ParLinearForm int_sigmoid_psi(psi.ParFESpace());
int_sigmoid_psi.AddDomainIntegrator(new DomainLFIntegrator(sigmoid_psi));
ParLinearForm int_der_sigmoid_psi(psi.ParFESpace());
int_der_sigmoid_psi.AddDomainIntegrator(new DomainLFIntegrator(
der_sigmoid_psi));
bool done = false;
for (int k=0; k<max_its; k++) // Newton iteration
{
int_sigmoid_psi.Assemble(); // Recompute f(c) with updated ψ
real_t f = int_sigmoid_psi.Sum();
MPI_Allreduce(MPI_IN_PLACE, &f, 1, MPITypeMap<real_t>::mpi_type,
MPI_SUM, MPI_COMM_WORLD);
f -= target_volume;
int_der_sigmoid_psi.Assemble(); // Recompute df(c) with updated ψ
real_t df = int_der_sigmoid_psi.Sum();
MPI_Allreduce(MPI_IN_PLACE, &df, 1, MPITypeMap<real_t>::mpi_type,
MPI_SUM, MPI_COMM_WORLD);
const real_t dc = -f/df;
psi += dc;
if (abs(dc) < tol) { done = true; break; }
}
if (!done)
{
mfem_warning("Projection reached maximum iteration without converging. "
"Result may not be accurate.");
}
int_sigmoid_psi.Assemble();
real_t material_volume = int_sigmoid_psi.Sum();
MPI_Allreduce(MPI_IN_PLACE, &material_volume, 1,
MPITypeMap<real_t>::mpi_type, MPI_SUM, MPI_COMM_WORLD);
return material_volume;
}
/*
* ---------------------------------------------------------------
* ALGORITHM PREAMBLE
@@ -138,11 +193,10 @@ int main(int argc, char *argv[])
int ref_levels = 5;
int order = 2;
real_t alpha = 1.0;
real_t growth = 2;
real_t epsilon = 0.01;
real_t vol_fraction = 0.5;
int max_it = 1e3;
real_t itol = 1e-2;
real_t itol = 1e-1;
real_t ntol = 1e-4;
real_t rho_min = 1e-6;
real_t lambda = 1.0;
@@ -157,8 +211,6 @@ int main(int argc, char *argv[])
"Order (degree) of the finite elements.");
args.AddOption(&alpha, "-alpha", "--alpha-step-length",
"Step length for gradient descent.");
args.AddOption(&growth, "-growth", "--alpha-growth-rate",
"Growth rate of step length for gradient descent.");
args.AddOption(&epsilon, "-epsilon", "--epsilon-thickness",
"Length scale for ρ.");
args.AddOption(&max_it, "-mi", "--max-it",
@@ -307,7 +359,6 @@ int main(int argc, char *argv[])
}
FilterSolver->SetEssentialBoundary(ess_bdr_filter);
FilterSolver->SetupFEM();
FilterSolver->AssembleDiffusionBilinear();
ParBilinearForm mass(&control_fes);
mass.AddDomainIntegrator(new InverseIntegrator(new MassIntegrator(one)));
@@ -361,7 +412,7 @@ int main(int argc, char *argv[])
// 11. Iterate:
for (int k = 1; k <= max_it; k++)
{
if (k > 1) { alpha = std::pow((real_t) k,growth); }
if (k > 1) { alpha *= ((real_t) k) / ((real_t) k-1); }
if (myid == 0)
{
@@ -401,9 +452,7 @@ int main(int argc, char *argv[])
// Step 5 - Update design variable ψ ← proj(ψ - αG)
psi.Add(-alpha, grad);
ParGridFunction alpha_grad(grad);
alpha_grad *= alpha;
const real_t material_volume = proj(psi, alpha_grad, target_volume);
const real_t material_volume = proj(psi, target_volume);
// Compute ||ρ - ρ_old|| in control fes.
real_t norm_increment = zerogf.ComputeL1Error(succ_diff_rho);
+1 -5
View File
@@ -9,7 +9,6 @@
// ex4 -m ../data/beam-hex.mesh -o 2 -pa
// ex4 -m ../data/escher.mesh
// ex4 -m ../data/fichera.mesh -o 2 -hb
// ex4 -m ../data/fichera.mesh -o 2 -hb -ea
// ex4 -m ../data/fichera-q2.vtk
// ex4 -m ../data/fichera-q3.mesh -o 2 -sc
// ex4 -m ../data/square-disc-nurbs.mesh
@@ -19,7 +18,6 @@
// ex4 -m ../data/amr-quad.mesh
// ex4 -m ../data/amr-hex.mesh
// ex4 -m ../data/amr-hex.mesh -o 2 -hb
// ex4 -m ../data/amr-hex.mesh -o 2 -hb -ea
// ex4 -m ../data/fichera-amr.mesh -o 2 -sc
// ex4 -m ../data/ref-prism.mesh -o 1
// ex4 -m ../data/octahedron.mesh -o 1
@@ -27,8 +25,6 @@
//
// Device sample runs:
// ex4 -m ../data/star.mesh -pa -d cuda
// ex4 -m ../data/star.mesh -hb -ea -d cuda
// ex4 -m ../data/amr-quad.mesh -hb -ea -d cuda
// ex4 -m ../data/star.mesh -pa -d raja-cuda
// ex4 -m ../data/star.mesh -pa -d raja-omp
// ex4 -m ../data/beam-hex.mesh -pa -d cuda
@@ -197,7 +193,7 @@ int main(int argc, char *argv[])
cout << "Size of linear system: " << A->Height() << endl;
// 11. Solve the linear system A X = B.
if (!pa && (!ea || hybridization))
if (!pa)
{
#ifndef MFEM_USE_SUITESPARSE
// Use a simple symmetric Gauss-Seidel preconditioner with PCG.
+1 -6
View File
@@ -9,7 +9,6 @@
// mpirun -np 4 ex4p -m ../data/beam-hex.mesh -o 2 -pa
// mpirun -np 4 ex4p -m ../data/escher.mesh -o 2 -sc
// mpirun -np 4 ex4p -m ../data/fichera.mesh -o 2 -hb
// mpirun -np 4 ex4p -m ../data/fichera.mesh -o 2 -hb -ea
// mpirun -np 4 ex4p -m ../data/fichera-q2.vtk
// mpirun -np 4 ex4p -m ../data/fichera-q3.mesh -o 2 -sc
// mpirun -np 4 ex4p -m ../data/square-disc-nurbs.mesh -o 3
@@ -18,18 +17,14 @@
// mpirun -np 4 ex4p -m ../data/periodic-cube.mesh -no-bc
// mpirun -np 4 ex4p -m ../data/amr-quad.mesh
// mpirun -np 3 ex4p -m ../data/amr-quad.mesh -o 2 -hb
// mpirun -np 3 ex4p -m ../data/amr-quad.mesh -o 2 -hb -ea
// mpirun -np 4 ex4p -m ../data/amr-hex.mesh -o 2 -sc
// mpirun -np 4 ex4p -m ../data/amr-hex.mesh -o 2 -hb
// mpirun -np 4 ex4p -m ../data/amr-hex.mesh -o 2 -hb -ea
// mpirun -np 4 ex4p -m ../data/ref-prism.mesh -o 1
// mpirun -np 4 ex4p -m ../data/octahedron.mesh -o 1
// mpirun -np 4 ex4p -m ../data/star-surf.mesh -o 3 -hb
//
// Device sample runs:
// mpirun -np 4 ex4p -m ../data/star.mesh -pa -d cuda
// mpirun -np 4 ex4p -m ../data/star.mesh -ea -hb -d cuda
// mpirun -np 4 ex4p -m ../data/amr-hex.mesh -ea -hb -d cuda
// mpirun -np 4 ex4p -m ../data/star.mesh -pa -d raja-cuda
// mpirun -np 4 ex4p -m ../data/star.mesh -pa -d raja-omp
// mpirun -np 4 ex4p -m ../data/beam-hex.mesh -pa -d cuda
@@ -235,7 +230,7 @@ int main(int argc, char *argv[])
pcg->SetMaxIter(2000);
pcg->SetPrintLevel(1);
if (hybridization) { prec = new HypreBoomerAMG(*A.As<HypreParMatrix>()); }
else if (pa || ea) { prec = new OperatorJacobiSmoother(*a, ess_tdof_list); }
else if (pa) { prec = new OperatorJacobiSmoother(*a, ess_tdof_list); }
else
{
ParFiniteElementSpace *prec_fespace =
+1 -1
View File
@@ -76,4 +76,4 @@ clean-build:
rm -rf *.dSYM *.TVD.*breakpoints
clean-exec:
@rm -f refined.mesh sol.gf mesh.* sol.*
@rm -f refined.mesh sol.gf
+2 -7
View File
@@ -71,7 +71,6 @@ endif
SUBDIRS_ALL = $(addsuffix /all,$(SUBDIRS))
SUBDIRS_TEST = $(addsuffix /test,$(SUBDIRS))
SUBDIRS_TEST_NOCLEAN = $(addsuffix /test-noclean,$(SUBDIRS))
SUBDIRS_CLEAN = $(addsuffix /clean,$(SUBDIRS))
SUBDIRS_TPRINT = $(addsuffix /test-print,$(SUBDIRS))
@@ -88,9 +87,8 @@ SUBDIRS_TPRINT = $(addsuffix /test-print,$(SUBDIRS))
all: $(EXAMPLES) $(SUBDIRS_ALL)
.PHONY: $(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) \
$(SUBDIRS_CLEAN) $(SUBDIRS_TPRINT)
$(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) $(SUBDIRS_CLEAN):
.PHONY: $(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_CLEAN) $(SUBDIRS_TPRINT)
$(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_CLEAN):
$(MAKE) -C $(@D) $(@F)
$(SUBDIRS_TPRINT):
@$(MAKE) -C $(@D) $(@F)
@@ -109,7 +107,6 @@ endif
MFEM_TESTS = EXAMPLES
include $(MFEM_TEST_MK)
test: $(SUBDIRS_TEST)
test-noclean: $(SUBDIRS_TEST_NOCLEAN)
test-print: $(SUBDIRS_TPRINT)
# Testing: Parallel vs. serial runs
@@ -160,8 +157,6 @@ ex37-test-seq: ex37
@$(call mfem-test,$<,, Serial example,-mi 3)
ex37p-test-par: ex37p
@$(call mfem-test,$<, $(RUN_MPI), Parallel example,-mi 3)
ex39-test-seq: ex39
@$(call mfem-test,$<,, Serial example,-m ../data/compass.mesh)
ex41-test-seq: ex41
@$(call mfem-test,$<,, Serial example,-tf 1.0)
ex41p-test-par: ex41p
+1 -8
View File
@@ -133,7 +133,7 @@ set(SRCS
tmop/assemble/diag2.cpp
tmop/assemble/grad2_limit.cpp
tmop/assemble/grad2.cpp
tmop/assemble/diag3_limit.cpp
tmop/assemble/diag3_limit.cpp
tmop/assemble/diag3.cpp
tmop/assemble/grad3_limit.cpp
tmop/assemble/grad3.cpp
@@ -171,12 +171,8 @@ set(SRCS
tmop_tools.cpp
tmop_amr.cpp
gslib.cpp
gslib/findptsedge_local_2.cpp
gslib/findptsedge_local_3.cpp
gslib/findptssurf_local_3.cpp
gslib/findpts_local_2.cpp
gslib/findpts_local_3.cpp
gslib/interpolate_local_1.cpp
gslib/interpolate_local_2.cpp
gslib/interpolate_local_3.cpp
transfer.cpp
@@ -195,14 +191,12 @@ set(HDRS
integ/bilininteg_dgtrace_kernels.hpp
integ/bilininteg_vecdiffusion_kernels.hpp
integ/bilininteg_convection_kernels.hpp
integ/bilininteg_diffusion_pa_simplices.hpp
integ/bilininteg_diffusion_kernels.hpp
integ/bilininteg_elasticity_kernels.hpp
integ/bilininteg_hcurl_kernels.hpp
integ/bilininteg_hdiv_kernels.hpp
integ/bilininteg_hcurlhdiv_kernels.hpp
integ/bilininteg_mass_kernels.hpp
integ/bilininteg_mass_pa_simplices.hpp
integ/bilininteg_vecdiffusion_pa.hpp
integ/bilininteg_vecmass_pa.hpp
coefficient.hpp
@@ -311,7 +305,6 @@ set(HDRS
tmop_tools.hpp
tmop_amr.hpp
gslib.hpp
gslib/gslib_kernel_helpers.hpp
transfer.hpp
hyperbolic.hpp
integrator.hpp
+9 -42
View File
@@ -729,8 +729,7 @@ void BilinearForm::Assemble(int skip_zeros)
tr = mesh -> GetBdrFaceTransformations (i);
if (tr != NULL)
{
mfem::DofTransformation doftrans;
fes -> GetElementVDofs (tr -> Elem1No, vdofs, doftrans);
fes -> GetElementVDofs (tr -> Elem1No, vdofs);
fe1 = fes -> GetFE (tr -> Elem1No);
// The fe2 object is really a dummy and not used on the boundaries,
// but we can't dereference a NULL pointer, and we don't want to
@@ -744,7 +743,6 @@ void BilinearForm::Assemble(int skip_zeros)
boundary_face_integs[k] -> AssembleFaceMatrix (*fe1, *fe2, *tr,
elemmat);
doftrans.TransformDual(elemmat);
mat -> AddSubMatrix (vdofs, vdofs, elemmat, skip_zeros);
}
}
@@ -827,46 +825,14 @@ void BilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list, Vector &x,
Vector &b, OperatorHandle &A, Vector &X,
Vector &B, int copy_interior)
{
const SparseMatrix *P = fes->GetConformingProlongation();
const SparseMatrix *R = fes->GetConformingRestriction();
if (ext)
{
if (hybridization)
{
FormSystemMatrix(ess_tdof_list, A);
std::unique_ptr<ConstrainedOperator> A_constrained([&]()
{
Operator *op;
Operator::FormSystemOperator(ess_tdof_list, op);
return dynamic_cast<ConstrainedOperator*>(op);
}());
MFEM_ASSERT(A_constrained != nullptr, "");
Vector conf_b, conf_x;
if (P)
{
// Nonconforming
conf_b.SetSize(P->Width());
conf_x.SetSize(P->Width());
P->MultTranspose(b, conf_b);
R->Mult(x, conf_x);
}
else
{
// Conforming
conf_b.MakeRef(b, 0, b.Size());
conf_x.MakeRef(x, 0, x.Size());
}
A_constrained->EliminateRHS(conf_x, conf_b);
if (P)
{
R->MultTranspose(conf_b, b); // store eliminated rhs in b
}
hybridization->ReduceRHS(conf_b, B);
ConstrainedOperator A_constrained(this, ess_tdof_list);
A_constrained.EliminateRHS(x, b);
hybridization->ReduceRHS(b, B);
X.SetSize(B.Size());
X = 0.0;
}
@@ -876,6 +842,7 @@ void BilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list, Vector &x,
}
return;
}
const SparseMatrix *P = fes->GetConformingProlongation();
FormSystemMatrix(ess_tdof_list, A);
// Transform the system and perform the elimination in B, based on the
@@ -911,6 +878,7 @@ void BilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list, Vector &x,
if (hybridization)
{
// Reduction to the Lagrange multipliers system
const SparseMatrix *R = fes->GetConformingRestriction();
Vector conf_b(P->Width()), conf_x(P->Width());
P->MultTranspose(b, conf_b);
R->Mult(x, conf_x);
@@ -923,6 +891,7 @@ void BilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list, Vector &x,
else
{
// Variational restriction with P
const SparseMatrix *R = fes->GetConformingRestriction();
B.SetSize(P->Width());
P->MultTranspose(b, B);
X.SetSize(R->Height());
@@ -1725,7 +1694,6 @@ void MixedBilinearForm::Assemble(int skip_zeros)
}
}
DofTransformation dom_dof_trans, ran_dof_trans;
for (int i = 0; i < trial_fes -> GetNBE(); i++)
{
const int bdr_attr = mesh->GetBdrAttribute(i);
@@ -1734,8 +1702,8 @@ void MixedBilinearForm::Assemble(int skip_zeros)
ftr = mesh -> GetBdrFaceTransformations (i);
if (ftr != NULL)
{
trial_fes->GetElementVDofs(ftr->Elem1No, trial_vdofs, dom_dof_trans);
test_fes->GetElementVDofs(ftr->Elem1No, test_vdofs, ran_dof_trans);
trial_fes->GetElementVDofs(ftr->Elem1No, trial_vdofs);
test_fes->GetElementVDofs(ftr->Elem1No, test_vdofs);
trial_fe1 = trial_fes->GetFE(ftr->Elem1No);
test_fe1 = test_fes->GetFE(ftr->Elem1No);
// The test_fe2 object is really a dummy and not used on the
@@ -1751,7 +1719,6 @@ void MixedBilinearForm::Assemble(int skip_zeros)
boundary_face_integs[k]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
*test_fe2,
*ftr, elemmat);
TransformDual(ran_dof_trans, dom_dof_trans, elemmat);
mat->AddSubMatrix(test_vdofs, trial_vdofs, elemmat, skip_zeros);
}
}
+4 -22
View File
@@ -1345,8 +1345,7 @@ real_t DiffusionIntegrator::ComputeFluxEnergy
}
const IntegrationRule &DiffusionIntegrator::GetRule(
const FiniteElement &trial_fe, const FiniteElement &test_fe,
const bool stroud)
const FiniteElement &trial_fe, const FiniteElement &test_fe)
{
int order;
if (trial_fe.Space() == FunctionSpace::Pk)
@@ -1363,15 +1362,7 @@ const IntegrationRule &DiffusionIntegrator::GetRule(
{
return RefinedIntRules.Get(trial_fe.GetGeomType(), order);
}
if (stroud)
{
return StroudIntRules.Get(trial_fe.GetGeomType(), order);
}
else
{
return IntRules.Get(trial_fe.GetGeomType(), order);
}
return IntRules.Get(trial_fe.GetGeomType(), order);
}
MassIntegrator::MassIntegrator(const IntegrationRule *ir)
@@ -1458,8 +1449,7 @@ void MassIntegrator::AssembleElementMatrix2(
const IntegrationRule &MassIntegrator::GetRule(const FiniteElement &trial_fe,
const FiniteElement &test_fe,
const ElementTransformation &Trans,
const bool stroud)
const ElementTransformation &Trans)
{
// int order = trial_fe.GetOrder() + test_fe.GetOrder();
const int order = trial_fe.GetOrder() + test_fe.GetOrder() + Trans.OrderW();
@@ -1468,15 +1458,7 @@ const IntegrationRule &MassIntegrator::GetRule(const FiniteElement &trial_fe,
{
return RefinedIntRules.Get(trial_fe.GetGeomType(), order);
}
if (stroud)
{
return StroudIntRules.Get(trial_fe.GetGeomType(), order);
}
else
{
return IntRules.Get(trial_fe.GetGeomType(), order);
}
return IntRules.Get(trial_fe.GetGeomType(), order);
}
+24 -55
View File
@@ -2137,14 +2137,14 @@ private:
int trial_dofs1D, test_dofs1D, quad1D;
public:
GradientIntegrator() :
Q{NULL}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
GradientIntegrator(const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q{NULL}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
{ }
GradientIntegrator(Coefficient *q_) :
Q{q_}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
GradientIntegrator(Coefficient *q_, const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q{q_}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
{ }
GradientIntegrator(Coefficient &q) :
Q{&q}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
GradientIntegrator(Coefficient &q, const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q{&q}, trial_maps{NULL}, test_maps{NULL}, geom{NULL}
{ }
void AssembleElementMatrix2(const FiniteElement &trial_fe,
@@ -2184,22 +2184,11 @@ public:
const Vector&, const Vector&,
Vector&, const int, const int);
using ApplySimplexKernelType = void(*)(const int, const bool, const Array<int>&,
const Array<int>&,
const Array<int>&, const Array<int>&, const Array<int>&,
const Array<real_t>&, const Array<real_t>&,
const Array<real_t>&, const Array<real_t>&,
const Array<real_t>&, const Array<real_t>&,
const Vector&, const Vector&,
Vector&, const int, const int);
using DiagonalKernelType = void(*)(const int, const bool, const Array<real_t>&,
const Array<real_t>&, const Vector&, Vector&,
const int, const int);
MFEM_REGISTER_KERNELS(ApplyPAKernels, ApplyKernelType, (int, int, int));
MFEM_REGISTER_KERNELS(ApplySimplexPAKernels, ApplySimplexKernelType, (int, int,
int));
MFEM_REGISTER_KERNELS(DiagonalPAKernels, DiagonalKernelType, (int, int, int));
struct Kernels { Kernels(); };
@@ -2352,8 +2341,7 @@ public:
void AddMultPatchPA(const int patch, const Vector &x, Vector &y) const;
static const IntegrationRule &GetRule(const FiniteElement &trial_fe,
const FiniteElement &test_fe,
const bool stroud = false);
const FiniteElement &test_fe);
bool SupportsCeed() const override { return DeviceCanUseCeed(); }
@@ -2364,13 +2352,6 @@ public:
{
ApplyPAKernels::Specialization<DIM,D1D,Q1D>::Add();
DiagonalPAKernels::Specialization<DIM,D1D,Q1D>::Add();
AddSimplexSpecialization<DIM,D1D,Q1D>();
}
template <int DIM, int D1D, int Q1D>
static void AddSimplexSpecialization()
{
ApplySimplexPAKernels::Specialization<DIM,D1D,Q1D>::Add();
}
protected:
const IntegrationRule* GetDefaultIntegrationRule(
@@ -2407,22 +2388,11 @@ public:
const Array<real_t>&, const Vector&,
const Vector&, Vector&, const int, const int);
using ApplySimplexKernelType = void(*)(const int, const Array<int>&,
const Array<int>&,
const Array<int>&, const Array<int>&, const Array<int>&,
const Array<real_t>&, const Array<real_t>&,
const Array<real_t>&, const Array<real_t>&,
const Array<real_t>&, const Array<real_t>&,
const Vector&, const Vector&, Vector&,
const int, const int);
using DiagonalKernelType = void(*)(const int, const Array<real_t>&,
const Vector&, Vector&, const int,
const int);
MFEM_REGISTER_KERNELS(ApplyPAKernels, ApplyKernelType, (int, int, int));
MFEM_REGISTER_KERNELS(ApplySimplexPAKernels, ApplySimplexKernelType, (int, int,
int));
MFEM_REGISTER_KERNELS(DiagonalPAKernels, DiagonalKernelType, (int, int, int));
struct Kernels { Kernels(); };
@@ -2471,8 +2441,7 @@ public:
static const IntegrationRule &GetRule(const FiniteElement &trial_fe,
const FiniteElement &test_fe,
const ElementTransformation &Trans,
const bool stroud = false);
const ElementTransformation &Trans);
bool SupportsCeed() const override { return DeviceCanUseCeed(); }
@@ -2483,13 +2452,6 @@ public:
{
ApplyPAKernels::Specialization<DIM,D1D,Q1D>::Add();
DiagonalPAKernels::Specialization<DIM,D1D,Q1D>::Add();
AddSimplexSpecialization<DIM,D1D,Q1D>();
}
template <int DIM, int D1D, int Q1D>
static void AddSimplexSpecialization()
{
ApplySimplexPAKernels::Specialization<DIM,D1D,Q1D>::Add();
}
protected:
@@ -2650,6 +2612,11 @@ protected:
Vector pa_data;
public:
VectorMassIntegrator(const IntegrationRule *ir)
: BilinearFormIntegrator(ir), vdim(-1), Q_order(0), Q(NULL), VQ(NULL),
MQ(NULL) { }
/// Construct an integrator with coefficient 1.0
VectorMassIntegrator() = default;
@@ -2748,7 +2715,7 @@ public:
/** Integrator for $(-Q u, \nabla v)$ for Nedelec ($u$) and $H^1$ ($v$) elements.
This is equivalent to a weak divergence of the $H(curl)$ basis functions. */
This is equivalent to a weak divergence of the $H(curl$ basis functions. */
class VectorFEWeakDivergenceIntegrator: public BilinearFormIntegrator
{
protected:
@@ -3076,14 +3043,15 @@ private:
int trial_dofs1D, test_dofs1D, quad1D;
public:
VectorDivergenceIntegrator() :
Q(NULL), trial_maps(NULL), test_maps(NULL), geom(NULL)
VectorDivergenceIntegrator(const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q(NULL), trial_maps(NULL), test_maps(NULL),
geom(NULL)
{ }
VectorDivergenceIntegrator(Coefficient *q_) :
Q(q_), trial_maps(NULL), test_maps(NULL), geom(NULL)
VectorDivergenceIntegrator(Coefficient *q_, const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q(q_), trial_maps(NULL), test_maps(NULL), geom(NULL)
{ }
VectorDivergenceIntegrator(Coefficient &q) :
Q(&q), trial_maps(NULL), test_maps(NULL), geom(NULL)
VectorDivergenceIntegrator(Coefficient &q, const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir), Q(&q), trial_maps(NULL), test_maps(NULL), geom(NULL)
{ }
void AssembleElementMatrix2(const FiniteElement &trial_fe,
@@ -3301,8 +3269,9 @@ private:
void SetUpQuadratureSpaceAndCoefficients(const FiniteElementSpace &fes);
public:
ElasticityIntegrator(Coefficient &l, Coefficient &m)
{ lambda = &l; mu = &m; }
ElasticityIntegrator(Coefficient &l, Coefficient &m,
const IntegrationRule *ir = NULL) :
BilinearFormIntegrator(ir) { lambda = &l; mu = &m; }
/** With this constructor $\lambda = q_l m$ and $\mu = q_m m$
if $dim q_l + 2 q_m = 0$ then $tr(\sigma) = 0$. */
ElasticityIntegrator(Coefficient &m, real_t q_l, real_t q_m)
+28 -67
View File
@@ -39,8 +39,8 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
b_type = b_type_i;
cp_type = cp_type_i;
tol = tol_i;
lbound.SetSize(ncp, nb);
ubound.SetSize(ncp, nb);
lbound.SetSize(nb, ncp);
ubound.SetSize(nb, ncp);
nodes.SetSize(nb);
weights.SetSize(nb);
control_points.SetSize(ncp);
@@ -125,25 +125,21 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
{
if (j == 0)
{
lbound(j,i) = bv(i);
ubound(j,i) = bv(i);
lbound(i, j) = bv(i);
ubound(i, j) = bv(i);
}
else if (j == ncp-1)
{
lbound(j,i) = bv(i);
ubound(j,i) = bv(i);
lbound(i, j) = bv(i);
ubound(i, j) = bv(i);
}
else
{
vals(0) = bv(i);
vals(1) = bmv(i) + dm*bdmv(i);
vals(2) = bpv(i) + dp*bdpv(i);
lbound(j,i) = vals.Min()-tol; // tolerance for good measure
ubound(j,i) = vals.Max()+tol; // tolerance for good measure
if (b_type == 2)
{
lbound(j,i) = std::max(lbound(j,i),0_r);
}
lbound(i, j) = vals.Min()-tol; // tolerance for good measure
ubound(i, j) = vals.Max()+tol; // tolerance for good measure
}
}
}
@@ -277,7 +273,8 @@ void PLBound::Get1DBounds(const Vector &coeff, Vector &intmin,
intmax.SetSize(ncp);
intmin = 0.0;
intmax = 0.0;
Vector coeffm;
Vector coeffm(nb);
coeffm = 0.0;
real_t a0 = 0.0;
real_t a1 = 0.0;
@@ -305,8 +302,6 @@ void PLBound::Get1DBounds(const Vector &coeff, Vector &intmin,
// compute L2 projection for linear bases: a0 + a1*x
if (proj)
{
coeffm.SetSize(nb);
coeffm = 0.0;
for (int i = 0; i < nb; i++)
{
x = 2.0*nodes_int(i)-1;
@@ -347,8 +342,8 @@ void PLBound::Get1DBounds(const Vector &coeff, Vector &intmin,
real_t c = coeffm(i);
for (int j = 0; j < ncp; j++)
{
intmin(j) += min(lbound(j,i)*c, ubound(j,i)*c);
intmax(j) += max(lbound(j,i)*c, ubound(j,i)*c);
intmin(j) += min(lbound(i,j)*c, ubound(i,j)*c);
intmax(j) += max(lbound(i,j)*c, ubound(i,j)*c);
}
}
}
@@ -479,10 +474,10 @@ void PLBound::Get2DBounds(const Vector &coeff, Vector &intmin,
real_t w1 = intmaxT(id2++);
for (int k = 0; k < ncp; k++) // kth row
{
vals(0) = w0*lbound(k,j);
vals(1) = w0*ubound(k,j);
vals(2) = w1*lbound(k,j);
vals(3) = w1*ubound(k,j);
vals(0) = w0*lbound(j,k);
vals(1) = w0*ubound(j,k);
vals(2) = w1*lbound(j,k);
vals(3) = w1*ubound(j,k);
intmin(k*ncp+i) += vals.Min();
intmax(k*ncp+i) += vals.Max();
}
@@ -558,17 +553,17 @@ void PLBound::Get3DBounds(const Vector &coeff, Vector &intmin,
for (int i = 0; i < nb; i++)
{
x = 2.0*nodes(i)-1; // x-coordinate
minNodalVals(i) -= a0V(j) + a1V(j)*x;
maxNodalVals(i) -= a0V(j) + a1V(j)*x;
minBounds(i) -= a0V(j) + a1V(j)*x;
maxBounds(i) -= a0V(j) + a1V(j)*x;
}
// Compute Bernstein coefficients
LUFactors lu(basisMatLU.GetData(), lu_ip.GetData());
lu.Solve(nb, 1, minNodalVals.GetData());
lu.Solve(nb, 1, maxNodalVals.GetData());
lu.Solve(nb, 1, minBounds.GetData());
lu.Solve(nb, 1, maxBounds.GetData());
for (int i = 0; i < nb; i++)
{
intminT(i*ncp2+j) = minNodalVals(i);
intmaxT(i*ncp2+j) = maxNodalVals(i);
intminT(i*ncp2+j) = minBounds(i);
intmaxT(i*ncp2+j) = maxBounds(i);
}
}
}
@@ -622,10 +617,10 @@ void PLBound::Get3DBounds(const Vector &coeff, Vector &intmin,
real_t w1 = intmaxT(id2++);
for (int k = 0; k < ncp; k++) // kth slice
{
vals(0) = w0*lbound(k,j);
vals(1) = w0*ubound(k,j);
vals(2) = w1*lbound(k,j);
vals(3) = w1*ubound(k,j);
vals(0) = w0*lbound(j,k);
vals(1) = w0*ubound(j,k);
vals(2) = w1*lbound(j,k);
vals(3) = w1*ubound(j,k);
intmin(k*ncp2+i) += vals.Min();
intmax(k*ncp2+i) += vals.Max();
}
@@ -658,8 +653,7 @@ void PLBound::SetupBernsteinBasisMat(DenseMatrix &basisMat,
Vector &nodesBern) const
{
const int nbern = nodesBern.Size();
L2_SegmentElement el(nbern-1, 2);
// we use L2 to leverage lexicographic order
L2_SegmentElement el(nbern-1, 2); // we use L2 to leverage lexicographic order
Array<int> ordering = el.GetLexicographicOrdering();
basisMat.SetSize(nbern, nbern);
Vector shape(nbern);
@@ -672,39 +666,6 @@ void PLBound::SetupBernsteinBasisMat(DenseMatrix &basisMat,
}
}
DenseMatrix PLBound::GetBoundingMatrix(int dim, bool is_lower) const
{
if (dim > 1)
{
const int ncpd = static_cast<int>(std::pow(ncp, dim));
const int nbd = static_cast<int>(std::pow(nb, dim));
DenseMatrix boundND(ncpd, nbd);
Vector phimin, phimax, col;
Vector coeffs(nbd);
coeffs = 0.0;
for (int j = 0; j < nbd; j++)
{
coeffs(j) = 1.0;
boundND.GetColumnReference(j, col);
GetNDBounds(dim, coeffs, phimin, phimax);
col = is_lower ? phimin : phimax;
coeffs(j) = 0.0;
}
return boundND;
}
return is_lower ? lbound : ubound;
}
DenseMatrix PLBound::GetLowerBoundMatrix(int dim) const
{
return GetBoundingMatrix(dim, true);
}
DenseMatrix PLBound::GetUpperBoundMatrix(int dim) const
{
return GetBoundingMatrix(dim, false);
}
constexpr int PLBound::min_ncp_gl_x[2][11];
constexpr int PLBound::min_ncp_gll_x[2][11];
constexpr int PLBound::min_ncp_pos_x[2][11];
@@ -755,4 +716,4 @@ void PLBound::Print(std::ostream &outp) const
ubound.Print(outp);
}
}
}
+20 -71
View File
@@ -19,18 +19,14 @@ namespace mfem
{
/** @name Piecewise linear bounds of bases
\brief Piecewise linear bounds of bases can be used to compute bounds on
the grid function in each element. The bounds for the bases are constructed
based on the following parameters:
\brief Piecewise linear bounds of bases can be used to compute bounds on the grid function in each element. The bounds for the bases are constructed based on the following parameters:
(i) @b nb: number of bases/nodes in 1D (i.e. polynomial order+1),
(ii) @b b_type: bases type, 0 - Lagrange interpolants on Gauss-Legendre
nodes, 1 - Lagrange interpolants on Gauss-Lobatto-Legendre nodes, and
(ii) @b b_type: bases type, 0 - Lagrange interpolants on Gauss-Legendre nodes, 1 - Lagrange interpolants on Gauss-Lobatto-Legendre nodes, and
2 - Positive/Bernstein bases on uniformly distributed nodes,
(iii) @b ncp: number of control points used to construct the piecewise
linear bounds
(iii) @b ncp: number of control points used to construct the piecewise linear bounds
(iv) @b cp_type: control point distribution. 0 - GL + end-points,
1 - Chebyshev.
@@ -39,9 +35,7 @@ namespace mfem
If the user does not specify @b ncp and @b cp_type, the minimum value of
@b ncp is used that would bound the bases for the @b cp_type. We default
to @b cp_type = 0 as it requires fewer number of points to bound the bases.
Typically, @b ncp = 2 @b nb is sufficient to get fairly compact bounds, and
increasing @b ncp results in tighter bounds.
to @b cp_type = 0 as it requires fewer number of points to bound the bases. Typically, @b ncp = 2 @b nb is sufficient to get fairly compact bounds, and increasing @b ncp results in tighter bounds.
Finally, only tensor-product elements are currently supported.
@@ -60,7 +54,7 @@ private:
bool proj = true; // Use linear projection to compute bounds.
real_t tol = 0.0; // offset bounds to avoid round-off errors
Vector nodes, weights, control_points;
DenseMatrix lbound, ubound; // ncp x nb matrices with bounds of all bases
DenseMatrix lbound, ubound; // nb x ncp matrices with bounds of all bases
// Some auxillary storage for computing the bounds with Bernstein
DenseMatrix basisMatNodes; // Bernstein bases at equispaced nodes
DenseMatrix basisMatInt; // Bernstein bases at GLL nodes
@@ -86,9 +80,6 @@ private:
{3,5,8,9,11,12,13,13,14,15,16}
};
/// Helper function to extract lower or upper bounding matrix
DenseMatrix GetBoundingMatrix(int dim, bool is_lower) const;
public:
// Constructor
PLBound(const int nb_i, const int ncp_i, const int b_type_i,
@@ -101,82 +92,40 @@ public:
PLBound(const FiniteElementSpace *fes,
const int ncp_i = -1, const int cp_type_i = 0);
/// Get minimum number of control points needed to bound the given bases
// Get minimum number of control points needed to bound the given bases
int GetMinimumPointsForGivenBases(int nb_i, int b_type_i,
int cp_type_i) const;
/// Print information about the bounds
// Print information about the bounds
void Print(std::ostream &outp = mfem::out) const;
/** @brief Enable (default) or disable linear projection before bounding.
*
* @details This projection increases the computational cost but results in
* tighter bounds.
*/
// Enable (default) or disable linear projection before bounding.
// This projection increases the computational cost but results in tighter
// bounds.
void SetProjectionFlagForBounding(bool proj_) { proj = proj_; }
/** @brief Compute piecewise linear bounds for the lexicographically-ordered
* nodal coefficients in @a coeff in 1D/2D/3D.
*
* @param[in] rdim The spatial dimension of the element (1, 2, or 3).
* @param[in] coeff The vector of lexicographically-ordered coefficients.
* Should be of size nb^rdim, where nb is the number of
* bases/nodes in 1D. These coefficients must correspond
* to the bases type and number of bases, used in the
* constructor of PLBound.
*
* @param[out] intmin The vector of minimum bound for all control points.
* @param[out] intmax The vector of maximum bound for all control points.
* Both intmin and intmax are of size ncp^rdim, where
* ncp is the number of control points in 1D, and are
* ordered lexicographically.
*/
/// Compute piecewise linear bounds for the lexicographically-ordered
/// coefficients in @a coeff in 1D/2D/3D.
void GetNDBounds(const int rdim, const Vector &coeff,
Vector &intmin, Vector &intmax) const;
/// Get number of control points used to compute the bounds.
int GetNControlPoints() const { return ncp; }
/// Get 1D control point locations (lexicographic order) in [0,1].
const Vector &GetControlPoints() const { return control_points; }
/** @brief Get lower and upper bounding matrix (ncp^dim x nb^dim)
*
* @details The matrices can be used to compute the bounds at control points
* by a simple matrix-vector product with the
* lexicographically-ordered nodal coefficients.
* The resulting output is also lexicographically-ordered.
*
* @note These matrices do not account for the linear projection step that
* is optionally done in GetNDBounds before bounding the function.
*/
///@{
DenseMatrix GetLowerBoundMatrix(int dim = 1) const;
DenseMatrix GetUpperBoundMatrix(int dim = 1) const;
///@}
private:
/** @brief Compute piecewise linear bounds for the lexicographically-ordered
* nodal coefficients in @a coeff in 1D.
* See GetNDBounds for details of the input and output parameters.
*/
/// Compute piecewise linear bounds for the lexicographically-ordered
/// coefficients in @a coeff in 1D.
void Get1DBounds(const Vector &coeff, Vector &intmin, Vector &intmax) const;
/** @brief Compute piecewise linear bounds for the lexicographically-ordered
* nodal coefficients in @a coeff in 2D.
* See GetNDBounds for details of the input and output parameters.
*/
/// Compute piecewise linear bounds for the lexicographically-ordered
/// coefficients in @a coeff in 2D.
void Get2DBounds(const Vector &coeff, Vector &intmin, Vector &intmax) const;
/** @brief Compute piecewise linear bounds for the lexicographically-ordered
* nodal coefficients in @a coeff in 3D.
* See GetNDBounds for details of the input and output parameters.
*/
/// Compute piecewise linear bounds for the lexicographically-ordered
/// coefficients in @a coeff in 3D.
void Get3DBounds(const Vector &coeff, Vector &intmin, Vector &intmax) const;
/** @brief Setup matrix used to compute values at given 1D locations in [0,1]
* for Bernstein bases.
*/
/// Setup matrix used to compute values at given 1D locations in [0,1]
/// for Bernstein bases.
void SetupBernsteinBasisMat(DenseMatrix &basisMat, Vector &nodesBern) const;
void Setup(const int nb_i, const int ncp_i, const int b_type_i,
-73
View File
@@ -54,8 +54,6 @@ void Coefficient::Project(QuadratureFunction &qf)
QuadratureSpaceBase &qspace = *qf.GetSpace();
const int ne = qspace.GetNE();
Vector values;
// GetValues makes a reference, but we need it to be valid on Host
qf.HostWrite();
for (int iel = 0; iel < ne; ++iel)
{
qf.GetValues(iel, values);
@@ -329,8 +327,6 @@ void VectorCoefficient::Project(QuadratureFunction &qf)
const int ne = qspace.GetNE();
DenseMatrix values;
Vector col;
// GetValues makes a reference, but we need it to be valid on Host
qf.HostWrite();
for (int iel = 0; iel < ne; ++iel)
{
qf.GetValues(iel, values);
@@ -699,8 +695,6 @@ void MatrixCoefficient::Project(QuadratureFunction &qf, bool transpose)
QuadratureSpaceBase &qspace = *qf.GetSpace();
const int ne = qspace.GetNE();
DenseMatrix values, matrix;
// GetValues makes a reference, but we need it to be valid on Host
qf.HostWrite();
for (int iel = 0; iel < ne; ++iel)
{
qf.GetValues(iel, values);
@@ -1308,73 +1302,6 @@ real_t TraceCoefficient::Eval(ElementTransformation &T,
return ma.Trace();
}
VectorComponentCoefficient::VectorComponentCoefficient(VectorCoefficient &A,
int c)
: a(&A), va(A.GetVDim())
{
SetComponent(c);
}
void VectorComponentCoefficient::SetComponent(int c)
{
MFEM_ASSERT(c < a->GetVDim() && c >= 0,
"VectorComponentCoefficient: "
"Index not in range.");
component = c;
}
void VectorComponentCoefficient::SetTime(real_t t)
{
if (a) { a->SetTime(t); }
this->Coefficient::SetTime(t);
}
real_t VectorComponentCoefficient::Eval(ElementTransformation &T,
const IntegrationPoint &ip)
{
a->Eval(va, T, ip);
return va[component];
}
MatrixComponentCoefficient::MatrixComponentCoefficient(MatrixCoefficient &A,
int ri, int ci)
: a(&A), ma(A.GetHeight(), A.GetWidth())
{
SetRowIndex(ri);
SetColumnIndex(ci);
}
void MatrixComponentCoefficient::SetRowIndex(int ri)
{
MFEM_ASSERT(ri < a->GetHeight() && ri >= 0,
"MatrixComponentCoefficient: "
"Row index not in range.");
row_idx = ri;
}
void MatrixComponentCoefficient::SetColumnIndex(int ci)
{
MFEM_ASSERT(ci < a->GetWidth() && ci >= 0,
"MatrixComponentCoefficient: "
"Column index not in range.");
col_idx = ci;
}
void MatrixComponentCoefficient::SetTime(real_t t)
{
if (a) { a->SetTime(t); }
this->Coefficient::SetTime(t);
}
real_t MatrixComponentCoefficient::Eval(ElementTransformation &T,
const IntegrationPoint &ip)
{
a->Eval(ma, T, ip);
return ma(row_idx,col_idx);
}
VectorSumCoefficient::VectorSumCoefficient(int dim)
: VectorCoefficient(dim),
ACoef(NULL), BCoef(NULL),
+5 -86
View File
@@ -52,9 +52,6 @@ public:
/// Get the time for time dependent coefficients
real_t GetTime() { return time; }
/// Returns dimension of the vector.
int GetVDim() { return 1; }
/** @brief Evaluate the coefficient in the element described by @a T at the
point @a ip. */
/** @note When this method is called, the caller must make sure that the
@@ -117,10 +114,11 @@ public:
/// Construct the constant coefficient using a vector of constants.
/** @a c should be a vector defined by attributes, so for region with
attribute @a i @a c[i-1] is the coefficient in that region */
PWConstCoefficient(const Vector &c) { UpdateConstants(c); }
PWConstCoefficient(Vector &c)
{ constants.SetSize(c.Size()); constants=c; }
/// Update the constants with vector @a c.
void UpdateConstants(const Vector &c) { constants = c; }
void UpdateConstants(Vector &c) { constants.SetSize(c.Size()); constants=c; }
/// Return a reference to the i-th constant
real_t &operator()(int i) { return constants(i-1); }
@@ -1334,8 +1332,8 @@ public:
/// Get the coefficient located at (i,j) in the matrix.
Coefficient* GetCoeff (int i, int j) { return Coeff[i*width+j]; }
/** @brief Set the coefficient located at (i,j) in the matrix. By default
this will take ownership of the Coefficient passed in, but this
/** @brief Set the coefficient located at (i,j) in the matrix. By default by
default this will take ownership of the Coefficient passed in, but this
can be overridden with the @a own parameter. */
void Set(int i, int j, Coefficient * c, bool own=true);
@@ -1875,85 +1873,6 @@ public:
const IntegrationPoint &ip) override;
};
/// Scalar coefficient defined as component of a vector coefficient
class VectorComponentCoefficient : public Coefficient
{
private:
VectorCoefficient *a = nullptr;
mutable Vector va;
int component;
public:
/// Construct with a vector coefficient.
VectorComponentCoefficient(VectorCoefficient &A)
: a(&A), va(A.GetVDim()), component(0) {};
VectorComponentCoefficient(VectorCoefficient &A, int c);
/// Set the time for internally stored coefficients
void SetTime(real_t t) override;
/// Reset the vector coefficient
void SetACoef(VectorCoefficient &A) { a = &A; }
/// Return the vector coefficient
VectorCoefficient * GetACoef() const { return a; }
/// Set the component
void SetComponent(int c);
/// Return the component
int GetComponent() const { return component; }
/// Evaluate the trace coefficient at @a ip.
real_t Eval(ElementTransformation &T,
const IntegrationPoint &ip) override;
};
/// Scalar coefficient defined as component of a matrix coefficient
class MatrixComponentCoefficient : public Coefficient
{
private:
MatrixCoefficient *a = nullptr;
mutable DenseMatrix ma;
int row_idx,col_idx;
public:
MatrixComponentCoefficient(MatrixCoefficient &A)
: a(&A), ma(A.GetHeight(), A.GetWidth()), row_idx(0), col_idx(0) {};
/// Construct with the matrix coefficient.
MatrixComponentCoefficient(MatrixCoefficient &A, int ri, int ci);
/// Set the time for internally stored coefficients
void SetTime(real_t t) override;
/// Reset the matrix coefficient
void SetACoef(MatrixCoefficient &A) { a = &A; }
/// Return the matrix coefficient
MatrixCoefficient * GetACoef() const { return a; }
/// Reset the index
void SetRowIndex(int ri);
/// Return the index
int GetRowIndex() const { return row_idx; }
/// Reset the index
void SetColumnIndex(int ci);
/// Return the index
int GetColumnIndex() const { return col_idx; }
/// Evaluate the trace coefficient at @a ip.
real_t Eval(ElementTransformation &T,
const IntegrationPoint &ip) override;
};
/// Vector coefficient defined as the linear combination of two vectors
class VectorSumCoefficient : public VectorCoefficient
{
+26 -88
View File
@@ -237,81 +237,6 @@ ComplexGridFunction::ProjectBdrCoefficientTangent(VectorCoefficient
gfi->SyncAliasMemory(*this);
}
real_t
ComplexGridFunction::ComputeLpError(const real_t p,
Coefficient &exsolr,
Coefficient &exsoli,
Coefficient *weight,
const IntegrationRule *irs[],
const Array<int> *elems) const
{
real_t error = 0.0;
const FiniteElement *fe;
ElementTransformation *T;
Vector valsr;
Vector valsi;
const GridFunction& gf_r = real();
const GridFunction& gf_i = imag();
for (int i = 0; i < fes->GetNE(); i++)
{
if (elems != NULL && (*elems)[i] == 0) { continue; }
fe = fes->GetFE(i);
const IntegrationRule *ir;
if (irs)
{
ir = irs[fe->GetGeomType()];
}
else
{
int intorder = 2*fe->GetOrder() + 3;
ir = &(IntRules.Get(fe->GetGeomType(), intorder));
}
real_t elem_error = 0.0;
gf_r.GetValues(i, *ir, valsr);
gf_i.GetValues(i, *ir, valsi);
T = fes->GetElementTransformation(i);
for (int j = 0; j < ir->GetNPoints(); j++)
{
const IntegrationPoint &ip = ir->IntPoint(j);
T->SetIntPoint(&ip);
real_t diffr = valsr(j) - exsolr.Eval(*T, ip);
real_t diffi = valsi(j) - exsoli.Eval(*T, ip);
real_t diff = hypot(diffr, diffi);
if (p < infinity())
{
diff = pow(diff, p);
if (weight)
{
diff *= weight->Eval(*T, ip);
}
elem_error += ip.weight * T->Weight() * diff;
}
else
{
if (weight)
{
diff *= weight->Eval(*T, ip);
}
error = std::max(error, diff);
}
}
if (p < infinity())
{
// negative quadrature weights may cause the error to be negative
error += fabs(elem_error);
}
}
if (p < infinity())
{
error = pow(error, 1./p);
}
return error;
}
void ComplexGridFunction::Save(std::ostream &os) const
{
os << "ComplexGridFunction\n";
@@ -905,9 +830,15 @@ ParComplexGridFunction::ParComplexGridFunction(ParMesh *m, std::istream &input)
int vsize = pfes->GetVSize();
Vector::Load(input, 2*vsize);
real_t *h_data = HostReadWrite();
pfes->ApplyDofSigns(h_data);
pfes->ApplyDofSigns(h_data + vsize);
real_t *data_ = const_cast<real_t*>(HostRead());
for (int i = 0; i < vsize; i++)
{
if (pfes->GetDofSign(i) < 0)
{
data_[i] = -data_[i];
data_[i+vsize] = -data_[i+vsize];
}
}
// if the mesh is a legacy (v1.1) NC mesh, it has old vertex ordering
@@ -1120,14 +1051,15 @@ void ParComplexGridFunction::Save(std::ostream &os) const
os << '\n';
int vsize = pfes->GetVSize();
// We use const_cast + HostRead (instead of HostReadWrite) because we only
// need to change the host data temporarily and this way we do not invalidate
// the data if it is on device. If we use HostReadWrite here, later calls to
// Read or ReadWrite will need to copy the data from host to device. With the
// approach used here, the host-to-device copy is avoided.
real_t *h_data = const_cast<real_t*>(HostRead());
pfes->ApplyDofSigns(h_data);
pfes->ApplyDofSigns(h_data + vsize);
real_t *data_ = const_cast<real_t*>(HostRead());
for (int i = 0; i < vsize; i++)
{
if (pfes->GetDofSign(i) < 0)
{
data_[i] = -data_[i];
data_[i+vsize] = -data_[i+vsize];
}
}
if (pfes->GetOrdering() == Ordering::byNODES)
{
@@ -1138,8 +1070,14 @@ void ParComplexGridFunction::Save(std::ostream &os) const
Vector::Print(os, pfes->GetVDim());
}
pfes->ApplyDofSigns(h_data);
pfes->ApplyDofSigns(h_data + vsize);
for (int i = 0; i < vsize; i++)
{
if (pfes->GetDofSign(i) < 0)
{
data_[i] = -data_[i];
data_[i+vsize] = -data_[i+vsize];
}
}
os.flush();
}
-88
View File
@@ -82,25 +82,6 @@ public:
/// underlying #fes
int VectorDim() const;
/// Copy assignment. Only the data of the base class Vector is copied.
/** It is assumed that this object and @a rhs use FiniteElementSpace%s that
have the same size.
@note Defining this method overwrites the implicitly defined copy
assignment operator. */
ComplexGridFunction &operator=(const ComplexGridFunction &rhs)
{ return operator=((const Vector &)rhs); }
/// Copy the data from @a v.
/** The size of @a v must be equal to double of the size of the associated
FiniteElementSpace #fes. */
ComplexGridFunction &operator=(const Vector &v)
{
MFEM_ASSERT(fes && v.Size() == 2*fes->GetVSize(), "");
Vector::operator=(v);
return *this;
}
/// Assign constant values to the ComplexGridFunction data.
ComplexGridFunction &operator=(const std::complex<real_t> & value)
{ *gfr = value.real(); *gfi = value.imag(); return *this; }
@@ -166,75 +147,6 @@ public:
return sqrt(err_r * err_r + err_i * err_i);
}
/// @brief Returns Max|u_ex - u_h| error for complex-valued H1 or L2 elements
///
/// Compute the $L_\infty$ error across the entire domain.
///
/// @param[in] exsolr Coefficient object reproducing the real part of the
/// anticipated values of the scalar field, Re(u_ex).
/// @param[in] exsoli Coefficient object reproducing the imaginary part of
/// the anticipated values of the scalar field, Im(u_ex).
/// @param[in] irs Optional pointer to an array of custom integration
/// rules e.g. higher order than the default rules. If
/// present the array will be indexed by
/// Geometry::Type.
///
/// @note Uses ComputeLpError internally. See the ComputeLpError
/// documentation for generalizations of this error computation.
///
/// @note If an array of integration rules is provided through @a irs, be
/// sure to include valid rules for each element type that may occur
/// in the list of elements.
///
virtual real_t ComputeMaxError(Coefficient &exsolr,
Coefficient &exsoli,
const IntegrationRule *irs[] = NULL) const
{
return ComputeLpError(infinity(), exsolr, exsoli, NULL, irs);
}
/// @brief Returns ||u_ex - u_h||_Lp for complex-valued H1 or L2 elements
///
/// Computes:
/// $$(\sum_{elems} \int_{elem} w \, |u_{ex} - u_h|^p)^{1/p}$$
/// Where:
/// $$|u_{ex} - u_h| = \sqrt{Re(u_{ex} - u_h)^2 + Im(u_{ex} - u_h)^2}$$
///
/// @param[in] p Real value indicating the exponent of the $L^p$ norm.
/// To avoid domain errors p should have a positive value,
/// either finite or infinite.
/// @param[in] exsolr Coefficient object reproducing the real part of the
/// anticipated values of the scalar field, Re(u_ex).
/// @param[in] exsoli Coefficient object reproducing the imaginary part of
/// the anticipated values of the scalar field, Im(u_ex).
/// @param[in] weight Optional pointer to a Coefficient object reproducing
/// a weighting function, w.
/// @param[in] irs Optional pointer to an array of custom integration
/// rules e.g. higher order than the default rules. If
/// present the array will be indexed by Geometry::Type.
/// @param[in] elems Optional pointer to a marker array, with a length
/// equal to the number of local elements, indicating
/// which elements to integrate over. Only those elements
/// corresponding to non-zero entries in @a elems will
/// contribute to the computed L2 error.
///
/// @note If an array of integration rules is provided through @a irs, be
/// sure to include valid rules for each element type that may occur
/// in the list of elements.
///
/// @note Quadratures with negative weights (as in some simplex integration
/// rules in MFEM) can produce negative integrals even with
/// non-negative integrands. To avoid returning negative errors this
/// function uses the absolute values of the element-wise integrals.
/// This may lead to results which are not entirely consistent with
/// such integration rules.
virtual real_t ComputeLpError(const real_t p,
Coefficient &exsolr,
Coefficient &exsoli,
Coefficient *weight = NULL,
const IntegrationRule *irs[] = NULL,
const Array<int> *elems = NULL) const;
/// Save the ComplexGridFunction to an output stream.
virtual void Save(std::ostream &out) const;
-4
View File
@@ -114,10 +114,6 @@ void ConduitDataCollection::Save()
n_mesh["fields"][name]);
}
// TODO: in parallel, we need to call ParFiniteElementSpace::ApplyDofSigns
// for all ParGridFunction objects before and after saving, see
// ParGridFunction::Save.
// save mesh data
SaveMeshAndFields(myid,
n_mesh,
+6 -21
View File
@@ -492,8 +492,6 @@ void VisItDataCollection::SaveRootFile()
to_padded_string(cycle, pad_digits_cycle) +
".mfem_root";
std::ofstream root_file(root_name);
MFEM_VERIFY(root_file.is_open(),
"Failed to open ofstream " << root_name);
root_file << GetVisItRootString();
if (!root_file)
{
@@ -979,10 +977,7 @@ void ParaViewDataCollection::Save()
// Save the local part of the mesh and grid functions fields to the local
// VTU file. Also save coefficient fields.
{
std::string os_str = vtu_prefix + GenerateVTUFileName("proc", myid);
std::ofstream os(os_str);
MFEM_VERIFY(os.is_open(),
"Failed to open ofstream " << os_str);
std::ofstream os(vtu_prefix + GenerateVTUFileName("proc", myid));
os.precision(precision);
SaveDataVTU(os, levels_of_detail);
}
@@ -994,10 +989,7 @@ void ParaViewDataCollection::Save()
"QuadratureFunction output is not supported for "
"ParaViewDataCollection on domain boundary!");
const std::string &field_name = qfield.first;
std::string os_str = vtu_prefix + GenerateVTUFileName(field_name, myid);
std::ofstream os(os_str);
MFEM_VERIFY(os.is_open(),
"Failed to open ofstream " << os_str);
std::ofstream os(vtu_prefix + GenerateVTUFileName(field_name, myid));
qfield.second->SaveVTU(os, pv_data_format, GetCompressionLevel(), field_name);
}
@@ -1008,10 +1000,7 @@ void ParaViewDataCollection::Save()
{
// Create the main PVTU file
{
std::string os_str = vtu_prefix + GeneratePVTUFileName("data");
std::ofstream pvtu_out(os_str);
MFEM_VERIFY(pvtu_out.is_open(),
"Failed to open ofstream " << os_str);
std::ofstream pvtu_out(vtu_prefix + GeneratePVTUFileName("data"));
WritePVTUHeader(pvtu_out);
// Grid function fields and coefficient fields
@@ -1066,10 +1055,8 @@ void ParaViewDataCollection::Save()
const std::string &q_field_name = q_field.first;
std::string q_fname = GeneratePVTUPath() + "/"
+ GeneratePVTUFileName(q_field_name);
std::string os_str = col_path + "/" + q_fname;
std::ofstream pvtu_out(os_str);
MFEM_VERIFY(pvtu_out.is_open(),
"Failed to open ofstream " << os_str);
std::ofstream pvtu_out(col_path + "/" + q_fname);
WritePVTUHeader(pvtu_out);
int vec_dim = q_field.second->GetVDim();
pvtu_out << "<PPointData>\n";
@@ -1181,14 +1168,12 @@ void ParaViewDataCollection::SaveGFieldVTU(std::ostream &os, int ref_,
DenseMatrix vval, pmat;
std::vector<char> buf;
int vec_dim = it->second->VectorDim();
int map_type = it->second->FESpace()->GetTypicalFE()->GetMapType();
os << "<DataArray type=\"" << GetDataTypeString()
<< "\" Name=\"" << it->first
<< "\" NumberOfComponents=\"" << vec_dim << "\" "
<< VTKComponentLabels(vec_dim) << " "
<< "format=\"" << GetDataFormatString() << "\" >" << '\n';
if (vec_dim == 1 && (map_type == FiniteElement::VALUE ||
map_type == FiniteElement::INTEGRAL))
if (vec_dim == 1)
{
for (int i = 0; i < mesh->GetNE(); i++)
{
+30 -79
View File
@@ -25,35 +25,21 @@
namespace mfem
{
/// Lightweight adaptor over an std::map from type K to type to V
template<typename K, typename V,
typename = typename std::enable_if<std::is_default_constructible<V>::value>::type>
class GenericFieldMap
/// Lightweight adaptor over an std::map from strings to pointer to T
template<typename T>
class NamedFieldsMap
{
private:
static constexpr bool ValueIsPointer = std::is_pointer<V>::value;
public:
typedef std::map<K, V> MapType;
typedef std::map<std::string, T*> MapType;
typedef typename MapType::iterator iterator;
typedef typename MapType::const_iterator const_iterator;
/// Register field @a field with name @a key
/// Only enabled if the template parameter V is not a pointer
template<typename = std::enable_if<!ValueIsPointer, bool>>
void Register(const K& key, V field)
/// Register field @a field with name @a fname
/** Replace existing field associated with @a fname (and optionally
delete associated pointer if @a own_data is true) */
void Register(const std::string& fname, T* field, bool own_data)
{
field_map[key] = field;
}
/// Register field @a field with name @a key
/** Replace existing field associated with @a key (and optionally
delete associated pointer if @a own_data is true).
Only enabled if the template parameter V is a pointer*/
template<typename = std::enable_if<ValueIsPointer, bool>>
void Register(const K& key, V field, bool own_data)
{
V& ref = field_map[key];
T*& ref = field_map[fname];
if (own_data)
{
delete ref; // if newly allocated -> ref is null -> OK
@@ -61,40 +47,23 @@ public:
ref = field;
}
/// Unregister association between field @a field and name @a key
/// Only enabled if the template parameter V is not a pointer
template<typename = std::enable_if<!ValueIsPointer, bool>>
void Deregister(const K& key)
/// Unregister association between field @a field and name @a fname
/** Optionally delete associated pointer if @a own_data is true */
void Deregister(const std::string& fname, bool own_data)
{
iterator it = field_map.find(key);
if ( it != field_map.end() )
{
field_map.erase(it);
}
}
/// Unregister association between field @a field and name @a key
/** Optionally delete associated pointer if @a own_data is true.
Only enabled if the template parameter V is a pointer */
template<typename = std::enable_if<ValueIsPointer, bool>>
void Deregister(const K& key, bool own_data)
{
iterator it = field_map.find(key);
iterator it = field_map.find(fname);
if ( it != field_map.end() )
{
if (own_data)
{
delete it->second;
it->second = nullptr;
}
field_map.erase(it);
}
}
/// Clear all associations between names and fields
/** Delete associated pointers when @a own_data is true.
Only enabled if the template parameter V is a pointer */
template<typename = std::enable_if<ValueIsPointer, bool>>
/** Delete associated pointers when @a own_data is true */
void DeleteData(bool own_data)
{
for (iterator it = field_map.begin(); it != field_map.end(); ++it)
@@ -107,37 +76,22 @@ public:
}
}
/// Predicate to check if a field is associated with name @a key
bool Has(const K& key) const
/// Predicate to check if a field is associated with name @a fname
bool Has(const std::string& fname) const
{
return field_map.find(key) != field_map.end();
return field_map.find(fname) != field_map.end();
}
/// Get a pointer to the field associated with name @a key
/** @return Field associated with @a key or NULL,
if value is pointer and key not found */
V Get(const K& key) const
/// Get a pointer to the field associated with name @a fname
/** @return Pointer to field associated with @a fname or NULL */
T* Get(const std::string& fname) const
{
const_iterator it = field_map.find(key);
if (it != field_map.end())
{
return it->second;
}
else
{
if constexpr (ValueIsPointer)
{
return nullptr;
}
else
{
return V(); // Return default-constructed value for non-pointer types
}
}
const_iterator it = field_map.find(fname);
return it != field_map.end() ? it->second : NULL;
}
/// Returns a const reference to the underlying map
const MapType &GetMap() const { return field_map; }
const MapType& GetMap() const { return field_map; }
/// Returns the number of registered fields
int NumFields() const { return field_map.size(); }
@@ -152,24 +106,21 @@ public:
/// Returns an end const iterator to the registered fields
const_iterator end() const { return field_map.end(); }
/// Returns an iterator to the field @a key
iterator find(const K& key)
{ return field_map.find(key); }
/// Returns an iterator to the field @a fname
iterator find(const std::string& fname)
{ return field_map.find(fname); }
/// Returns a const iterator to the field @a key
const_iterator find(const K& key) const
{ return field_map.find(key); }
/// Returns a const iterator to the field @a fname
const_iterator find(const std::string& fname) const
{ return field_map.find(fname); }
/// Clears the map of registered fields
/// Clears the map of registered fields without reclaiming memory
void clear() { field_map.clear(); }
protected:
MapType field_map;
};
/// Lightweight adaptor over an std::map from strings to pointer to T
template<typename T>
using NamedFieldsMap = GenericFieldMap<std::string, T*>;
/** A class for collecting finite element data that is part of the same
simulation. Currently, this class groups together grid functions (fields),
+8 -11
View File
@@ -90,8 +90,8 @@ void map_quadrature_data_to_fields_impl(
}
else
{
MFEM_ABORT_KERNEL("quadrature data mapping to field is not implemented"
" for this field descriptor");
MFEM_ABORT("quadrature data mapping to field is not implemented for"
" this field descriptor");
}
}
@@ -169,9 +169,8 @@ void map_quadrature_data_to_fields_tensor_impl_1d(
}
else
{
MFEM_ABORT_KERNEL("quadrature data mapping to field is not implemented"
"for this field descriptor with sum factorization on"
" tensor product elements");
MFEM_ABORT("quadrature data mapping to field is not implemented for"
" this field descriptor with sum factorization on tensor product elements");
}
}
@@ -307,9 +306,8 @@ void map_quadrature_data_to_fields_tensor_impl_2d(
}
else
{
MFEM_ABORT_KERNEL("quadrature data mapping to field is not implemented"
" for this field descriptor with sum factorization on"
" tensor product elements");
MFEM_ABORT("quadrature data mapping to field is not implemented for"
" this field descriptor with sum factorization on tensor product elements");
}
}
@@ -494,9 +492,8 @@ void map_quadrature_data_to_fields_tensor_impl_3d(
}
else
{
MFEM_ABORT_KERNEL("quadrature data mapping to field is not implemented"
" for this field descriptor with sum factorization on"
" tensor product elements");
MFEM_ABORT("quadrature data mapping to field is not implemented for"
" this field descriptor with sum factorization on tensor product elements");
}
}
+14 -6
View File
@@ -652,6 +652,16 @@ void forall(func_t f,
class FDJacobian : public Operator
{
public:
FDJacobian(const Operator &op, real_t fixed_eps = 0.0) :
Operator(op.Height(), op.Width()),
op(op),
x(op.Width()),
fixed_eps(fixed_eps)
{
f.SetSize(Height());
xpev.SetSize(Width());
}
FDJacobian(const Operator &op, const Vector &x, real_t fixed_eps = 0.0) :
Operator(op.Height(), op.Width()),
op(op),
@@ -671,9 +681,9 @@ public:
MPI_COMM_WORLD);
}
Operator& GetGradient(const Vector &x0) const override
void Update(const Vector &x_new)
{
x = x0;
x = x_new;
f.UseDevice(x.UseDevice());
xpev.UseDevice(x.UseDevice());
@@ -681,8 +691,6 @@ public:
const real_t xnorm_local = x.Norml2();
MPI_Allreduce(&xnorm_local, &xnorm, 1, MPITypeMap<real_t>::mpi_type, MPI_SUM,
MPI_COMM_WORLD);
return const_cast<FDJacobian&>(*this);
}
void Mult(const Vector &v, Vector &y) const override
@@ -739,11 +747,11 @@ public:
private:
const Operator &op;
mutable Vector x, f;
Vector x, f;
mutable Vector xpev;
real_t lambda = 1.0e-6;
real_t fixed_eps;
mutable real_t xnorm;
real_t xnorm;
};
/// @brief Find the index of a field descriptor in a vector of field descriptors.
+165
View File
@@ -12,6 +12,7 @@
#include "dgmassinv.hpp"
#include "bilinearform.hpp"
#include "dgmassinv_kernels.hpp"
#include "../general/forall.hpp"
namespace mfem
{
@@ -118,6 +119,151 @@ void DGMassInverse::Update()
DGMassInverse::~DGMassInverse() = default;
template<int DIM, int D1D, int Q1D>
void DGMassInverse::DGMassCGIteration(const Vector &b_, Vector &u_) const
{
using namespace internal; // host/device kernel functions
const int NE = fes.GetNE();
const int d1d = m->dofs1D;
const int q1d = m->quad1D;
const int ND = static_cast<int>(pow(d1d, DIM));
const auto B = m->maps->B.Read();
const auto Bt = m->maps->Bt.Read();
const auto pa_data = m->pa_data.Read();
const auto dinv = diag_inv.Read();
auto r = r_.Write();
auto d = d_.Write();
auto z = z_.Write();
auto u = u_.ReadWrite();
const real_t RELTOL = rel_tol;
const real_t ABSTOL = abs_tol;
const int MAXIT = max_iter;
const bool IT_MODE = iterative_mode;
const bool CHANGE_BASIS = (d2q != nullptr);
// b is the right-hand side (if no change of basis, this just points to the
// incoming RHS vector, if we have to change basis, this points to the
// internal b2 vector where we put the transformed RHS)
const real_t *b;
// the following are non-null if we have to change basis
real_t *b2 = nullptr; // non-const access to b2
const real_t *b_orig = nullptr; // RHS vector in "original" basis
const real_t *d2q_B = nullptr; // matrix to transform initial guess
const real_t *q2d_B = nullptr; // matrix to transform solution
const real_t *q2d_Bt = nullptr; // matrix to transform RHS
if (CHANGE_BASIS)
{
d2q_B = d2q->B.Read();
q2d_B = B_.Read();
q2d_Bt = Bt_.Read();
b2 = b2_.Write();
b_orig = b_.Read();
b = b2;
}
else
{
b = b_.Read();
}
static constexpr int NB = Q1D ? Q1D : 1; // block size
mfem::forall_2D(NE, NB, NB, [=] MFEM_HOST_DEVICE (int e)
{
// Perform change of basis if needed
if (CHANGE_BASIS)
{
// Transform RHS
DGMassBasis<DIM,D1D>(e, NE, q2d_Bt, b_orig, b2, d1d);
if (IT_MODE)
{
// Transform initial guess
DGMassBasis<DIM,D1D>(e, NE, d2q_B, u, u, d1d);
}
}
const int tid = MFEM_THREAD_ID(x) + NB*MFEM_THREAD_ID(y);
// Compute first residual
if (IT_MODE)
{
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, u, r, d1d, q1d);
DGMassAxpy(e, NE, ND, 1.0, b, -1.0, r, r); // r = b - r
}
else
{
// if not in iterative mode, use zero initial guess
const int BX = MFEM_THREAD_SIZE(x);
const int BY = MFEM_THREAD_SIZE(y);
const int bxy = BX*BY;
const auto B = ConstDeviceMatrix(b, ND, NE);
auto U = DeviceMatrix(u, ND, NE);
auto R = DeviceMatrix(r, ND, NE);
for (int i = tid; i < ND; i += bxy)
{
U(i, e) = 0.0;
R(i, e) = B(i, e);
}
MFEM_SYNC_THREAD;
}
DGMassPreconditioner(e, NE, ND, dinv, r, z);
DGMassAxpy(e, NE, ND, 1.0, z, 0.0, z, d); // d = z
real_t nom = DGMassDot<NB>(e, NE, ND, d, r);
if (nom < 0.0) { return; /* Not positive definite */ }
real_t r0 = fmax(nom*RELTOL*RELTOL, ABSTOL*ABSTOL);
if (nom <= r0) { return; /* Converged */ }
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, d, z, d1d, q1d);
real_t den = DGMassDot<NB>(e, NE, ND, z, d);
if (den <= 0.0)
{
DGMassDot<NB>(e, NE, ND, d, d);
// d2 > 0 => not positive definite
if (den == 0.0) { return; }
}
// start iteration
int i = 1;
while (true)
{
const real_t alpha = nom/den;
DGMassAxpy(e, NE, ND, 1.0, u, alpha, d, u); // u = u + alpha*d
DGMassAxpy(e, NE, ND, 1.0, r, -alpha, z, r); // r = r - alpha*A*d
DGMassPreconditioner(e, NE, ND, dinv, r, z);
real_t betanom = DGMassDot<NB>(e, NE, ND, r, z);
if (betanom < 0.0) { return; /* Not positive definite */ }
if (betanom <= r0) { break; /* Converged */ }
if (++i > MAXIT) { break; }
const real_t beta = betanom/nom;
DGMassAxpy(e, NE, ND, 1.0, z, beta, d, d); // d = z + beta*d
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, d, z, d1d, q1d); // z = A d
den = DGMassDot<NB>(e, NE, ND, d, z);
if (den <= 0.0)
{
DGMassDot<NB>(e, NE, ND, d, d);
// d2 > 0 => not positive definite
if (den == 0.0) { break; }
}
nom = betanom;
}
if (CHANGE_BASIS)
{
DGMassBasis<DIM,D1D>(e, NE, q2d_B, u, u, d1d);
}
});
}
void DGMassInverse::Mult(const Vector &Mu, Vector &u) const
{
// Dispatch to templated version based on dim, d1d, and q1d.
@@ -160,4 +306,23 @@ DGMassInvKernels::DGMassInvKernels()
k::Specialization<3,6,7>::Add();
}
/// @cond Suppress_Doxygen_warnings
template <int DIM, int D1D, int Q1D>
DGMassInverse::CGKernelType DGMassInverse::CGKernels::Kernel()
{
return &DGMassInverse::DGMassCGIteration<DIM,D1D,Q1D>;
}
DGMassInverse::CGKernelType DGMassInverse::CGKernels::Fallback(
int dim, int, int)
{
if (dim == 1) { return &DGMassInverse::DGMassCGIteration<1>; }
else if (dim == 2) { return &DGMassInverse::DGMassCGIteration<2>; }
else if (dim == 3) { return &DGMassInverse::DGMassCGIteration<3>; }
else { MFEM_ABORT("Unsupported dimension."); }
}
/// @endcond
} // namespace mfem
+1 -166
View File
@@ -15,7 +15,6 @@
#include "../linalg/kernels.hpp"
#include "kernels.hpp"
#include "integ/bilininteg_mass_kernels.hpp"
#include "dgmassinv.hpp"
namespace mfem
{
@@ -57,7 +56,7 @@ void DGMassApply(const int e,
}
else if (DIM == 3)
{
SmemPAMassApply3D_Element<TD1D,TQ1D,NBZ,ACCUM>(e, NE, B, pa_data, x, y);
SmemPAMassApply3D_Element<TD1D,TQ1D,ACCUM>(e, NE, B, pa_data, x, y);
}
else
{
@@ -334,170 +333,6 @@ void DGMassBasis(const int e,
} // namespace internal
template<int DIM, int D1D, int Q1D>
void DGMassInverse::DGMassCGIteration(const Vector &b_, Vector &u_) const
{
using namespace internal; // host/device kernel functions
const int NE = fes.GetNE();
const int d1d = m->dofs1D;
const int q1d = m->quad1D;
const int ND = static_cast<int>(pow(d1d, DIM));
const auto B = m->maps->B.Read();
const auto Bt = m->maps->Bt.Read();
const auto pa_data = m->pa_data.Read();
const auto dinv = diag_inv.Read();
auto r = r_.Write();
auto d = d_.Write();
auto z = z_.Write();
auto u = u_.ReadWrite();
const real_t RELTOL = rel_tol;
const real_t ABSTOL = abs_tol;
const int MAXIT = max_iter;
const bool IT_MODE = iterative_mode;
const bool CHANGE_BASIS = (d2q != nullptr);
// b is the right-hand side (if no change of basis, this just points to the
// incoming RHS vector, if we have to change basis, this points to the
// internal b2 vector where we put the transformed RHS)
const real_t *b;
// the following are non-null if we have to change basis
real_t *b2 = nullptr; // non-const access to b2
const real_t *b_orig = nullptr; // RHS vector in "original" basis
const real_t *d2q_B = nullptr; // matrix to transform initial guess
const real_t *q2d_B = nullptr; // matrix to transform solution
const real_t *q2d_Bt = nullptr; // matrix to transform RHS
if (CHANGE_BASIS)
{
d2q_B = d2q->B.Read();
q2d_B = B_.Read();
q2d_Bt = Bt_.Read();
b2 = b2_.Write();
b_orig = b_.Read();
b = b2;
}
else
{
b = b_.Read();
}
static constexpr int NB = Q1D ? Q1D : 1; // block size
mfem::forall_2D<NB*NB>(NE, NB, NB, [=] MFEM_HOST_DEVICE (int e)
{
// Perform change of basis if needed
if (CHANGE_BASIS)
{
// Transform RHS
DGMassBasis<DIM,D1D>(e, NE, q2d_Bt, b_orig, b2, d1d);
if (IT_MODE)
{
// Transform initial guess
DGMassBasis<DIM,D1D>(e, NE, d2q_B, u, u, d1d);
}
}
const int tid = MFEM_THREAD_ID(x) + NB*MFEM_THREAD_ID(y);
// Compute first residual
if (IT_MODE)
{
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, u, r, d1d, q1d);
DGMassAxpy(e, NE, ND, 1.0, b, -1.0, r, r); // r = b - r
}
else
{
// if not in iterative mode, use zero initial guess
const int BX = MFEM_THREAD_SIZE(x);
const int BY = MFEM_THREAD_SIZE(y);
const int bxy = BX*BY;
const auto B = ConstDeviceMatrix(b, ND, NE);
auto U = DeviceMatrix(u, ND, NE);
auto R = DeviceMatrix(r, ND, NE);
for (int i = tid; i < ND; i += bxy)
{
U(i, e) = 0.0;
R(i, e) = B(i, e);
}
MFEM_SYNC_THREAD;
}
DGMassPreconditioner(e, NE, ND, dinv, r, z);
DGMassAxpy(e, NE, ND, 1.0, z, 0.0, z, d); // d = z
real_t nom = DGMassDot<NB>(e, NE, ND, d, r);
if (nom < 0.0) { return; /* Not positive definite */ }
real_t r0 = fmax(nom*RELTOL*RELTOL, ABSTOL*ABSTOL);
if (nom <= r0) { return; /* Converged */ }
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, d, z, d1d, q1d);
real_t den = DGMassDot<NB>(e, NE, ND, z, d);
if (den <= 0.0)
{
DGMassDot<NB>(e, NE, ND, d, d);
// d2 > 0 => not positive definite
if (den == 0.0) { return; }
}
// start iteration
int i = 1;
while (true)
{
const real_t alpha = nom/den;
DGMassAxpy(e, NE, ND, 1.0, u, alpha, d, u); // u = u + alpha*d
DGMassAxpy(e, NE, ND, 1.0, r, -alpha, z, r); // r = r - alpha*A*d
DGMassPreconditioner(e, NE, ND, dinv, r, z);
real_t betanom = DGMassDot<NB>(e, NE, ND, r, z);
if (betanom < 0.0) { return; /* Not positive definite */ }
if (betanom <= r0) { break; /* Converged */ }
if (++i > MAXIT) { break; }
const real_t beta = betanom/nom;
DGMassAxpy(e, NE, ND, 1.0, z, beta, d, d); // d = z + beta*d
DGMassApply<DIM,D1D,Q1D>(e, NE, B, Bt, pa_data, d, z, d1d, q1d); // z = A d
den = DGMassDot<NB>(e, NE, ND, d, z);
if (den <= 0.0)
{
DGMassDot<NB>(e, NE, ND, d, d);
// d2 > 0 => not positive definite
if (den == 0.0) { break; }
}
nom = betanom;
}
if (CHANGE_BASIS)
{
DGMassBasis<DIM,D1D>(e, NE, q2d_B, u, u, d1d);
}
});
}
/// @cond Suppress_Doxygen_warnings
template <int DIM, int D1D, int Q1D>
inline DGMassInverse::CGKernelType DGMassInverse::CGKernels::Kernel()
{
return &DGMassInverse::DGMassCGIteration<DIM,D1D,Q1D>;
}
inline DGMassInverse::CGKernelType DGMassInverse::CGKernels::Fallback(
int dim, int, int)
{
if (dim == 1) { return &DGMassInverse::DGMassCGIteration<1>; }
else if (dim == 2) { return &DGMassInverse::DGMassCGIteration<2>; }
else if (dim == 3) { return &DGMassInverse::DGMassCGIteration<3>; }
else { MFEM_ABORT("Unsupported dimension."); }
}
/// @endcond
} // namespace mfem
#endif
+6 -6
View File
@@ -320,8 +320,8 @@ public:
error estimation procedure where the flux averaging is replaced by a global
L2 projection (requiring a mass matrix solve).
The required BilinearFormIntegrator must implement the method
ComputeElementFlux().
The required BilinearFormIntegrator must implement the methods
ComputeElementFlux() and ComputeFluxEnergy().
Implemented for the parallel case only.
*/
@@ -357,8 +357,8 @@ protected:
public:
/** @brief Construct a new L2ZienkiewiczZhuEstimator object.
@param integ This BilinearFormIntegrator must implement the method
ComputeElementFlux().
@param integ This BilinearFormIntegrator must implement the methods
ComputeElementFlux() and ComputeFluxEnergy().
@param sol The solution field whose error is to be estimated.
@param flux_fes The L2ZienkiewiczZhuEstimator assumes ownership of this
FiniteElementSpace and will call its Update() method when
@@ -382,8 +382,8 @@ public:
{ }
/** @brief Construct a new L2ZienkiewiczZhuEstimator object.
@param integ This BilinearFormIntegrator must implement the method
ComputeElementFlux().
@param integ This BilinearFormIntegrator must implement the methods
ComputeElementFlux() and ComputeFluxEnergy().
@param sol The solution field whose error is to be estimated.
@param flux_fes The L2ZienkiewiczZhuEstimator does NOT assume ownership
of this FiniteElementSpace; will call its Update() method
+3 -3
View File
@@ -69,9 +69,9 @@ inline int ToLexOrdering2D(const int face_id, const int size1d, const int i)
}
/// @brief Given a face DOF index on a shared face, ordered lexicographically
/// relative to the element (where the local face is face_id), return the
/// corresponding face DOF index ordered lexicographically relative to the face
/// itself.
/// relative to element the element (where the local face is face_id), and
/// return the corresponding face DOF index ordered lexicographically relative
/// to the face itself.
MFEM_HOST_DEVICE
inline int PermuteFace2D(const int face_id, const int orientation,
const int size1d, const int index)
+76 -142
View File
@@ -231,7 +231,7 @@ void FiniteElement::CalcPhysLaplacian(ElementTransformation &Trans,
{
for (int nd = 0; nd < dof; nd++)
{
Laplacian[nd] = hess(nd,0) + hess(nd,3) + hess(nd,5);
Laplacian[nd] = hess(nd,0) + hess(nd,4) + hess(nd,5);
}
}
else if (dim == 2)
@@ -268,9 +268,11 @@ void FiniteElement::CalcPhysLinLaplacian(ElementTransformation &Trans,
scale[0] = Gij(0,0);
scale[1] = 2*Gij(0,1);
scale[2] = 2*Gij(0,2);
scale[3] = Gij(1,1);
scale[4] = 2*Gij(1,2);
scale[5] = Gij(2,2);
scale[3] = 2*Gij(1,2);
scale[4] = Gij(2,2);
scale[5] = Gij(1,1);
}
else if (dim == 2)
{
@@ -307,12 +309,12 @@ void FiniteElement::CalcPhysHessian(ElementTransformation &Trans,
map[2] = 2;
map[3] = 1;
map[4] = 3;
map[5] = 4;
map[4] = 5;
map[5] = 3;
map[6] = 2;
map[7] = 4;
map[8] = 5;
map[7] = 3;
map[8] = 4;
}
else if (dim == 2)
{
@@ -380,7 +382,11 @@ const DofToQuad &FiniteElement::GetDofToQuad(const IntegrationRule &ir,
#pragma omp critical (DofToQuad)
#endif
{
d2q = DofToQuad::SearchArray(dof2quad_array, ir, mode);
for (int i = 0; i < dof2quad_array.Size(); i++)
{
d2q = dof2quad_array[i];
if (d2q->IntRule != &ir || d2q->mode != mode) { d2q = nullptr; }
}
if (!d2q)
{
#ifdef MFEM_THREAD_SAFE
@@ -655,67 +661,58 @@ void ScalarFiniteElement::ScalarLocalL2Restriction(
void NodalFiniteElement::CreateLexicographicFullMap(const IntegrationRule &ir)
const
{
// Get the FULL version of the map. This call contains omp critical region,
// so it is done before the critical region below.
auto &d2q = GetDofToQuad(ir, DofToQuad::FULL);
#if defined(MFEM_THREAD_SAFE) && defined(MFEM_USE_OPENMP)
#pragma omp critical (DofToQuad)
#endif
{
// Do not run if the new Dof2Quad is already present, e.g. added in a
// previous call or added by another omp thread.
if (DofToQuad::SearchArray(dof2quad_array, ir,
DofToQuad::LEXICOGRAPHIC_FULL) == nullptr)
// Get the FULL version of the map.
auto &d2q = GetDofToQuad(ir, DofToQuad::FULL);
//Undo the native ordering which is what FiniteElement::GetDofToQuad returns.
auto *d2q_new = new DofToQuad(d2q);
d2q_new->mode = DofToQuad::LEXICOGRAPHIC_FULL;
const int nqpt = ir.GetNPoints();
const int b_dim = (range_type == VECTOR) ? dim : 1;
for (int i = 0; i < nqpt; i++)
{
// Undo the native ordering which is what FiniteElement::GetDofToQuad
// returns.
auto *d2q_new = new DofToQuad(d2q);
d2q_new->mode = DofToQuad::LEXICOGRAPHIC_FULL;
const int nqpt = ir.GetNPoints();
const int b_dim = (range_type == VECTOR) ? dim : 1;
for (int i = 0; i < nqpt; i++)
for (int d = 0; d < b_dim; d++)
{
for (int d = 0; d < b_dim; d++)
for (int j = 0; j < dof; j++)
{
for (int j = 0; j < dof; j++)
{
const double val = d2q.B[i + nqpt*(d+b_dim*lex_ordering[j])];
d2q_new->B[i+nqpt*(d+b_dim*j)] = val;
d2q_new->Bt[j+dof*(i+nqpt*d)] = val;
}
const double val = d2q.B[i + nqpt*(d+b_dim*lex_ordering[j])];
d2q_new->B[i+nqpt*(d+b_dim*j)] = val;
d2q_new->Bt[j+dof*(i+nqpt*d)] = val;
}
}
const int g_dim = [this]()
{
switch (deriv_type)
{
case GRAD: return dim;
case DIV: return 1;
case CURL: return cdim;
default: return 0;
}
}();
for (int i = 0; i < nqpt; i++)
{
for (int d = 0; d < g_dim; d++)
{
for (int j = 0; j < dof; j++)
{
const double val = d2q.G[i + nqpt*(d+g_dim*lex_ordering[j])];
d2q_new->G[i+nqpt*(d+g_dim*j)] = val;
d2q_new->Gt[j+dof*(i+nqpt*d)] = val;
}
}
}
dof2quad_array.Append(d2q_new);
}
const int g_dim = [this]()
{
switch (deriv_type)
{
case GRAD: return dim;
case DIV: return 1;
case CURL: return cdim;
default: return 0;
}
}();
for (int i = 0; i < nqpt; i++)
{
for (int d = 0; d < g_dim; d++)
{
for (int j = 0; j < dof; j++)
{
const double val = d2q.G[i + nqpt*(d+g_dim*lex_ordering[j])];
d2q_new->G[i+nqpt*(d+g_dim*j)] = val;
d2q_new->Gt[j+dof*(i+nqpt*d)] = val;
}
}
}
dof2quad_array.Append(d2q_new);
}
}
@@ -727,7 +724,13 @@ const DofToQuad &NodalFiniteElement::GetDofToQuad(const IntegrationRule &ir,
#pragma omp critical (DofToQuad)
#endif
{
d2q = DofToQuad::SearchArray(dof2quad_array, ir, mode);
//Should make this loop a function of FiniteElement
for (int i = 0; i < dof2quad_array.Size(); i++)
{
d2q = dof2quad_array[i];
if (d2q->IntRule == &ir && d2q->mode == mode) { break; }
d2q = nullptr;
}
}
if (d2q) { return *d2q; }
if (mode != DofToQuad::LEXICOGRAPHIC_FULL)
@@ -1044,50 +1047,9 @@ void VectorFiniteElement::SetDerivMembers()
switch (map_type)
{
case H_DIV:
switch (dim)
{
case 3: // div: 3D H_DIV -> 3D INTEGRAL
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
case 2: // div: 2D H_DIV -> 2D INTEGRAL
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
break;
case H_DIV_R2D:
switch (dim)
{
case 2: // div: 2D H_DIV_R2D -> 2D INTEGRAL
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
case 1: // div: 1D H_DIV_R2D -> 1D INTEGRAL
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
break;
case H_DIV_R1D:
switch (dim)
{
case 1: // div: 1D H_DIV_R1D -> 1D INTEGRAL
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
deriv_type = DIV;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
case H_CURL:
switch (dim)
@@ -1105,49 +1067,13 @@ void VectorFiniteElement::SetDerivMembers()
break;
case 1:
deriv_type = NONE;
deriv_range_type = UNKNOWN_RANGE_TYPE;
deriv_map_type = UNKNOWN_MAP_TYPE;
deriv_range_type = SCALAR;
deriv_map_type = INTEGRAL;
break;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
break;
case H_CURL_R2D:
switch (dim)
{
case 2:
// curl: 2D H_CURL_R2D -> H_DIV_R2D
deriv_type = CURL;
deriv_range_type = VECTOR;
deriv_map_type = H_DIV_R2D;
break;
case 1:
// curl: 1D H_CURL_R2D -> H_DIV_R2D
deriv_type = CURL;
deriv_range_type = VECTOR;
deriv_map_type = H_DIV_R2D;
break;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
break;
case H_CURL_R1D:
switch (dim)
{
case 1:
// curl: 1D H_CURL_R1D -> H_DIV_R1D
deriv_type = CURL;
deriv_range_type = VECTOR;
deriv_map_type = H_DIV_R1D;
break;
case 0:
deriv_type = NONE;
deriv_range_type = UNKNOWN_RANGE_TYPE;
deriv_map_type = UNKNOWN_MAP_TYPE;
default:
MFEM_ABORT("Invalid dimension, Dim = " << dim);
}
break;
default:
MFEM_ABORT("Invalid MapType = " << map_type);
}
@@ -2705,7 +2631,15 @@ const DofToQuad &TensorBasisElement::GetTensorDofToQuad(
#pragma omp critical (DofToQuad)
#endif
{
d2q = DofToQuad::SearchArray(dof2quad_array, ir, mode);
for (int i = 0; i < dof2quad_array.Size(); i++)
{
auto* d2q_ = dof2quad_array[i];
if (d2q_->IntRule == &ir && d2q_->mode == mode)
{
d2q = d2q_;
break;
}
}
if (!d2q)
{
d2q = new DofToQuad;
+7 -98
View File
@@ -44,7 +44,7 @@ public:
NumBasisTypes = 9 /**< Keep track of maximum types to prevent
hard-coding */
};
/** @brief If the input does not represent a valid BasisType, abort with an
/** @brief If the input does not represents a valid BasisType, abort with an
error; otherwise return the input. */
static int Check(int b_type)
{
@@ -52,7 +52,7 @@ public:
"unknown BasisType: " << b_type);
return b_type;
}
/** @brief If the input does not represent a valid nodal BasisType, abort
/** @brief If the input does not represents a valid nodal BasisType, abort
with an error; otherwise return the input. */
static int CheckNodal(int b_type)
{
@@ -167,15 +167,7 @@ public:
/** @brief Full multidimensional representation which does not use tensor
product structure. The ordering of the degrees of freedom is the
same as TENSOR, but the sizes of B and G are the same as FULL.*/
LEXICOGRAPHIC_FULL,
/** @brief Ragged tensor product representation using 1D matrices/tensors
with dimensions using 1D number of quadrature points and ragged tensor degrees of
freedom. */
/** Used only for partial assembly of the H1 positive basis. The
size of B is d1d x qnpt x dim. Since different Gauss-Jacobi quadrature rules
are employed in each dimension, we need to store dim arrays. */
RAGGED_TENSOR
LEXICOGRAPHIC_FULL
};
/// Describes the contents of the #B, #Bt, #G, and #Gt arrays, see #Mode.
@@ -230,45 +222,6 @@ public:
/// Returns absolute value of the maps
DofToQuad Abs() const;
/// Auxiliary function for searching DofToQuad arrays.
static inline DofToQuad *SearchArray(
const Array<DofToQuad*> &dof2quad_array,
const IntegrationRule &ir,
DofToQuad::Mode mode);
virtual ~DofToQuad() = default;
};
/** @brief Structure representing the matrices/tensors needed to evaluate (in
reference space) the values, gradients, divergences, or curls of a positive
FiniteElement on simplices at the quadrature points of Stroud conical quadrature. */
class RaggedDofToQuad : public DofToQuad
{
public:
/** @brief Special basis function structures for positive (Bernstein) basis with
partial assembly. The storage layout of Ba1 is ndof x nqpt for scalar elements.
The storage layout of Ba2 is ndof x ndof x nqpt. In particular, we have
Ba2(iqpt, a1, a2) = B^{p-a1}_{a2}(x_{iqpt}). */
Array<real_t> Ba1, Ba2, Ba3;
Array<real_t> Ba1t, Ba2t, Ba3t;
/** @brief Special structures for gradients of positive basis with partial assembly.
The gradient arrays exploit properties of the Bernstein basis which allow grad(B^p_alpha)
to be expressed as the sum of products of B^{p-1}_alpha and the barycentric coordinates.
Thus, Ga1 and Ga2 simply contain the ragged tensor product components of B^{p-1}_alpha */
Array<real_t> Ga1, Ga2, Ga3;
Array<real_t> Ga1t, Ga2t, Ga3t;
/** @brief Mapping from the Bernstein multi-index (a_1, ..., a_d) to the lexicographic
dof index. */
Array<int> lex_map;
Array<int> forward_map2d_diff, forward_map3d_diff;
Array<int> inverse_map2d_diff, inverse_map3d_diff;
Array<int> forward_map2d_mass, forward_map3d_mass;
Array<int> inverse_map2d_mass, inverse_map3d_mass;
};
/// Describes the function space on each element
@@ -336,20 +289,10 @@ public:
$ u(x) = (1/w) \hat u(\hat x) $ */
H_DIV, /**< For vector fields; preserves surface integrals of the
normal component $ u(x) = (J/w) \hat u(\hat x) $ */
H_CURL, /**< For vector fields; preserves line integrals of the
H_CURL /**< For vector fields; preserves line integrals of the
tangential component
$ u(x) = J^{-t} \hat u(\hat x) $ (square J),
$ u(x) = J(J^t J)^{-1} \hat u(\hat x) $ (general J) */
H_DIV_R2D, /**< For 3-component vector fields in 2D; equivalent to a
direct sum of an H_DIV basis and an INTEGRAL basis */
H_CURL_R2D,/**< For 3-component vector fields in 2D; equivalent to a
direct sum of an H_CURL basis and a VALUE basis */
H_DIV_R1D, /**< For 3-component vector fields in 1D; equivalent to a
direct sum of a VALUE basis and a pair of INTEGRAL
bases */
H_CURL_R1D /**< For 3-component vector fields in 1D; equivalent to a
direct sum of an INTEGRAL basis and a pair of VALUE
bases */
};
/** @brief Enumeration for DerivType: defines which derivative method
@@ -381,28 +324,12 @@ public:
int GetDim() const { return dim; }
/** @brief Returns the vector dimension for vector-valued finite elements,
which is also the dimension of the interpolation operation and the
width of the DenseMatrix argument in
CalcVShape(const IntegrationPoint &ip, DenseMatrix &shape). */
which is also the dimension of the interpolation operation. */
int GetRangeDim() const { return vdim; }
/** @brief Returns the vector dimension, in physical space, for
vector-valued finite elements, which is also the width of the
DenseMatrix argument in
CalcPhysVShape(ElementTransformation &Trans, DenseMatrix &shape). */
virtual int GetPhysRangeDim(int /* space_dim */) const { return vdim; }
/** Returns the dimension of the curl for vector-valued finite elements,
which is also the width of the DenseMatrix argument in
CalcCurlShape(const IntegrationPoint &ip, DenseMatrix &curl_shape). */
/// Returns the dimension of the curl for vector-valued finite elements.
int GetCurlDim() const { return cdim; }
/** Returns the dimension, in physical space, of the curl for vector-valued
finite elements, which is also the width of the DenseMatrix argument in
CalcPhysCurlShape(ElementTransformation &Trans, DenseMatrix &curl_shape).
*/
virtual int GetPhysCurlDim(int /* space_dim */) const { return cdim; }
/// Returns the Geometry::Type of the reference element.
Geometry::Type GetGeomType() const { return geom_type; }
@@ -480,7 +407,6 @@ public:
/** Each row of the result DenseMatrix @a Hessian contains upper triangular
part of the Hessian of one shape function.
The order in 2D is {u_xx, u_xy, u_yy}.
The order in 3D is {u_xx, u_xy, u_xz, u_yy, u_yz, u_zz}.
The size (#dof x (#dim (#dim+1)/2) of @a Hessian must be set in advance.*/
virtual void CalcHessian(const IntegrationPoint &ip,
DenseMatrix &Hessian) const;
@@ -1057,8 +983,6 @@ protected:
public:
VectorFiniteElement(int D, Geometry::Type G, int Do, int O, int M,
int F = FunctionSpace::Pk);
int GetPhysRangeDim(int space_dim) const override { return space_dim; }
};
/// @brief Class for computing 1D special polynomials and their associated basis
@@ -1196,7 +1120,7 @@ public:
return GetPoints(p, btype, on_device);
}
/// Get coordinates of a closed (GaussLobatto) set of points if degree @a p
/// Get coordinates of a closed (GaussLegendre) set of points if degree @a p
const real_t *ClosedPoints(const int p,
const int btype = BasisType::GaussLobatto,
bool on_device = false)
@@ -1452,21 +1376,6 @@ public:
void InvertLinearTrans(ElementTransformation &trans,
const IntegrationPoint &pt, Vector &x);
// static inline method
inline DofToQuad *DofToQuad::SearchArray(
const Array<DofToQuad*> &dof2quad_array,
const IntegrationRule &ir,
DofToQuad::Mode mode)
{
for (int i = 0; i < dof2quad_array.Size(); i++)
{
DofToQuad *d2q = dof2quad_array[i];
if (d2q->IntRule == &ir && d2q->mode == mode) { return d2q; }
}
return nullptr;
}
} // namespace mfem
#endif
-48
View File
@@ -60,12 +60,6 @@ void Linear1DFiniteElement::CalcDShape(const IntegrationPoint &ip,
dshape(1,0) = 1.;
}
void Linear1DFiniteElement::CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const
{
h = 0.0;
}
Linear2DFiniteElement::Linear2DFiniteElement()
: NodalFiniteElement(2, Geometry::TRIANGLE, 3, 1)
{
@@ -93,11 +87,6 @@ void Linear2DFiniteElement::CalcDShape(const IntegrationPoint &ip,
dshape(2,0) = 0.; dshape(2,1) = 1.;
}
void Linear2DFiniteElement::CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const
{
h = 0.0;
}
BiLinear2DFiniteElement::BiLinear2DFiniteElement()
: NodalFiniteElement(2, Geometry::SQUARE, 4, 1, FunctionSpace::Qk)
@@ -1267,12 +1256,6 @@ void Linear3DFiniteElement::CalcDShape(const IntegrationPoint &ip,
}
}
void Linear3DFiniteElement::CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const
{
h = 0.0;
}
void Linear3DFiniteElement::GetFaceDofs (int face, int **dofs, int *ndofs)
const
{
@@ -1649,37 +1632,6 @@ void TriLinear3DFiniteElement::CalcDShape(const IntegrationPoint &ip,
dshape(7,2) = ox * y;
}
void TriLinear3DFiniteElement::CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const
{
real_t x = ip.x, y = ip.y, z = ip.z;
real_t ox = 1.-x, oy = 1.-y, oz = 1.-z;
h(0,0) = 0.; h(0,1) = oz; h(0,2) = oy;
h(0,3) = 0.; h(0,4) = ox; h(0,5) = 0.;
h(1,0) = 0.; h(1,1) = -oz; h(1,2) = -oy;
h(1,3) = 0.; h(1,4) = x; h(1,5) = 0.;
h(2,0) = 0.; h(2,1) = oz; h(2,2) = -y;
h(2,3) = 0.; h(2,4) = -x; h(2,5) = 0.;
h(3,0) = 0.; h(3,1) = -oz; h(3,2) = y;
h(3,3) = 0.; h(3,4) = -ox; h(3,5) = 0.;
h(4,0) = 0.; h(4,1) = z; h(4,2) = -oy;
h(4,3) = 0.; h(4,4) = -ox; h(4,5) = 0.;
h(5,0) = 0.; h(5,1) = -z; h(5,2) = oy;
h(5,3) = 0.; h(5,4) = -x; h(5,5) = 0.;
h(6,0) = 0.; h(6,1) = z; h(6,2) = y;
h(6,3) = 0.; h(6,4) = x; h(6,5) = 0.;
h(7,0) = 0.; h(7,1) = -z; h(7,2) = -y;
h(7,3) = 0.; h(7,4) = ox; h(7,5) = 0.;
}
P0SegmentFiniteElement::P0SegmentFiniteElement(int Ord)
: NodalFiniteElement(1, Geometry::SEGMENT, 1, Ord) // default Ord = 0
+1 -9
View File
@@ -50,8 +50,6 @@ public:
contains the derivative of one shape function */
void CalcDShape(const IntegrationPoint &ip,
DenseMatrix &dshape) const override;
void CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const override;
};
/// A 2D linear element on triangle with nodes at the vertices of the triangle
@@ -72,8 +70,6 @@ public:
so that each row contains the derivatives of one shape function */
void CalcDShape(const IntegrationPoint &ip,
DenseMatrix &dshape) const override;
void CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const override;
void ProjectDelta(int vertex, Vector &dofs) const override
{ dofs = 0.0; dofs(vertex) = 1.0; }
};
@@ -408,9 +404,6 @@ public:
void CalcDShape(const IntegrationPoint &ip,
DenseMatrix &dshape) const override;
void CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const override;
void ProjectDelta(int vertex, Vector &dofs) const override
{ dofs = 0.0; dofs(vertex) = 1.0; }
@@ -452,8 +445,7 @@ public:
so that each row contains the derivatives of one shape function */
void CalcDShape(const IntegrationPoint &ip,
DenseMatrix &dshape) const override;
void CalcHessian(const IntegrationPoint &ip,
DenseMatrix &h) const override;
void ProjectDelta(int vertex, Vector &dofs) const override
{ dofs = 0.0; dofs(vertex) = 1.0; }
};
+1 -1
View File
@@ -589,7 +589,7 @@ void H1_TriangleElement::CalcHessian(const IntegrationPoint &ip,
Vector shape_x(p + 1), shape_y(p + 1), shape_l(p + 1);
Vector dshape_x(p + 1), dshape_y(p + 1), dshape_l(p + 1);
Vector ddshape_x(p + 1), ddshape_y(p + 1), ddshape_l(p + 1);
DenseMatrix ddu(dof, (dim*(dim+1))/2);
DenseMatrix ddu(dof, dim);
#endif
poly1d.CalcBasis(p, ip.x, shape_x, dshape_x, ddshape_x);
+4 -4
View File
@@ -2531,7 +2531,7 @@ void ND_FuentesPyramidElement::calcCurlBasis(const int p,
ND_R1D_PointElement::ND_R1D_PointElement(int p)
: VectorFiniteElement(1, Geometry::POINT, 2, p,
H_CURL_R1D, FunctionSpace::Pk)
H_CURL, FunctionSpace::Pk)
{
// VectorFiniteElement::SetDerivMembers doesn't support 0D H_CURL elements
// so we mimic a 1D element and then correct the dimension here.
@@ -2562,7 +2562,7 @@ ND_R1D_SegmentElement::ND_R1D_SegmentElement(const int p,
const int cb_type,
const int ob_type)
: VectorFiniteElement(1, Geometry::SEGMENT, 3 * p + 2, p,
H_CURL_R1D, FunctionSpace::Pk),
H_CURL, FunctionSpace::Pk),
dof2tk(dof),
cbasis1d(poly1d.GetBasis(p, VerifyClosed(cb_type))),
obasis1d(poly1d.GetBasis(p - 1, VerifyOpen(ob_type)))
@@ -2839,7 +2839,7 @@ ND_R2D_SegmentElement::ND_R2D_SegmentElement(const int p,
const int cb_type,
const int ob_type)
: VectorFiniteElement(1, Geometry::SEGMENT, 2 * p + 1, p,
H_CURL_R2D, FunctionSpace::Pk),
H_CURL, FunctionSpace::Pk),
dof2tk(dof),
cbasis1d(poly1d.GetBasis(p, VerifyClosed(cb_type))),
obasis1d(poly1d.GetBasis(p - 1, VerifyOpen(ob_type)))
@@ -3023,7 +3023,7 @@ void ND_R2D_SegmentElement::Project(VectorCoefficient &vc,
ND_R2D_FiniteElement::ND_R2D_FiniteElement(int p, Geometry::Type G, int Do,
const real_t *tk_fe)
: VectorFiniteElement(2, G, Do, p,
H_CURL_R2D, FunctionSpace::Pk),
H_CURL, FunctionSpace::Pk),
tk(tk_fe),
dof_map(dof),
dof2tk(dof)
-6
View File
@@ -663,9 +663,6 @@ public:
const int cb_type = BasisType::GaussLobatto,
const int ob_type = BasisType::GaussLegendre);
int GetPhysRangeDim(int space_dim) const override { return 2; }
int GetPhysCurlDim(int space_dim) const override { return 1; }
void CalcVShape(const IntegrationPoint &ip,
DenseMatrix &shape) const override;
@@ -708,9 +705,6 @@ private:
DenseMatrix &I) const;
public:
int GetPhysRangeDim(int space_dim) const override { return 3; }
int GetPhysCurlDim(int space_dim) const override { return 3; }
using FiniteElement::CalcVShape;
using FiniteElement::CalcPhysCurlShape;
+5 -519
View File
@@ -84,46 +84,6 @@ void NURBS1DFiniteElement::CalcHessian (const IntegrationPoint &ip,
add(1.0, hess, (-d2sum + 2*dsum*dsum*sum)*sum*sum, shape_x, hess);
}
void NURBS1DFiniteElement::Project(Coefficient &coeff,
ElementTransformation &Trans,
Vector &dofs) const
{
IntegrationPoint ip;
for (int i = 0; i <= order; i++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+order)) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+order);
Trans.SetIntPoint(&ip);
dofs(i) = coeff.Eval(Trans, ip);
}
}
void NURBS1DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == vc.GetVDim()*dof, "");
Vector x(vc.GetVDim());
IntegrationPoint ip;
for (int i = 0; i <= order; i++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+order)) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+order);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
for (int j = 0; j < x.Size(); j++)
{
dofs(dof*j+i) = x(j);
}
}
}
void NURBS2DFiniteElement::SetOrder() const
{
@@ -255,63 +215,6 @@ void NURBS2DFiniteElement::CalcHessian (const IntegrationPoint &ip,
}
}
void NURBS2DFiniteElement::Project(Coefficient &coeff,
ElementTransformation &Trans,
Vector &dofs) const
{
IntegrationPoint ip;
for (int o = 0, j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 1;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
dofs(o) = coeff.Eval(Trans, ip);
}
}
}
void NURBS2DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == vc.GetVDim()*dof, "");
Vector x(vc.GetVDim());
IntegrationPoint ip;
for (int o = 0, j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 1;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
for (int v = 0; v < x.Size(); v++)
{
dofs(dof*v+o) = x(v);
}
}
}
}
void NURBS3DFiniteElement::SetOrder() const
{
@@ -445,10 +348,11 @@ void NURBS3DFiniteElement::CalcHessian (const IntegrationPoint &ip,
d2sum[0] += ( hessian(o,0) = d2sx*sy*sz*weights(o) );
d2sum[1] += ( hessian(o,1) = dsx*dsy*sz*weights(o) );
d2sum[2] += ( hessian(o,2) = dsx*sy*dsz*weights(o) );
d2sum[3] += ( hessian(o,3) = sx*d2sy*sz*weights(o) );
d2sum[4] += ( hessian(o,4) = sx*dsy*dsz*weights(o) );
d2sum[5] += ( hessian(o,5) = sx*sy*d2sz*weights(o) );
d2sum[3] += ( hessian(o,3) = sx*dsy*dsz*weights(o) );
d2sum[4] += ( hessian(o,4) = sx*sy*d2sz*weights(o) );
d2sum[5] += ( hessian(o,5) = sx*d2sy*sz*weights(o) );
}
}
}
@@ -497,85 +401,6 @@ void NURBS3DFiniteElement::CalcHessian (const IntegrationPoint &ip,
}
}
void NURBS3DFiniteElement::Project(Coefficient &coeff,
ElementTransformation &Trans,
Vector &dofs) const
{
IntegrationPoint ip;
for (int o = 0, k = 0; k <= orders[2]; k++)
{
real_t kz = kv[2]->GetBotella(ijk[2] + k);
if (!kv[2]->inSpan(kz, ijk[2]+orders[2]))
{
o += (orders[0] + 1)*(orders[1] + 1);
continue;
}
ip.z = kv[2]->GetRefPoint(kz, ijk[2]+orders[2]);
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 1;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
dofs(o) = coeff.Eval(Trans, ip);
}
}
}
}
void NURBS3DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == vc.GetVDim()*dof, "");
Vector x(vc.GetVDim());
IntegrationPoint ip;
for (int o = 0, k = 0; k <= orders[2]; k++)
{
real_t kz = kv[2]->GetBotella(ijk[2] + k);
if (!kv[2]->inSpan(kz, ijk[2]+orders[2]))
{
o += (orders[0] + 1)*(orders[1] + 1);
continue;
}
ip.z = kv[2]->GetRefPoint(kz, ijk[2]+orders[2]);
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 1;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
for (int v = 0; v < x.Size(); v++)
{
dofs(dof*v+o) = x(v);
}
}
}
}
}
void NURBS_HDiv2DFiniteElement::SetOrder() const
{
@@ -692,63 +517,6 @@ void NURBS_HDiv2DFiniteElement::CalcDivShape(const IntegrationPoint &ip,
}
}
void NURBS_HDiv2DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == dof, "");
MFEM_ASSERT(vc.GetVDim() == 2, "");
Vector x(2), mx(2);
IntegrationPoint ip;
int o = 0;
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 2;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]+1; i++, o++)
{
real_t kx = kv1[0]->GetBotella(ijk[0] + i);
if (!kv1[0]->inSpan(kx, ijk[0]+orders[0]+1)) { continue; }
ip.x = kv1[0]->GetRefPoint(kx, ijk[0]+orders[0]+1);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.AdjugateJacobian().Mult(x,mx);
dofs(o) = mx(0);
}
}
for (int j = 0; j <= orders[1]+1; j++)
{
real_t ky = kv1[1]->GetBotella(ijk[1] + j);
if (!kv1[1]->inSpan(ky, ijk[1]+orders[1]+1))
{
o += orders[0] + 1;
continue;
}
ip.y = kv1[1]->GetRefPoint(ky, ijk[1]+orders[1]+1);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.AdjugateJacobian().Mult(x,mx);
dofs(o) = mx(1);
}
}
}
NURBS_HDiv2DFiniteElement::~NURBS_HDiv2DFiniteElement()
{
if (kv1[0]) { delete kv1[0]; }
@@ -928,120 +696,6 @@ void NURBS_HDiv3DFiniteElement::CalcDivShape(const IntegrationPoint &ip,
}
}
void NURBS_HDiv3DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == dof, "");
MFEM_ASSERT(vc.GetVDim() == 3, "");
Vector x(2), mx(3);
IntegrationPoint ip;
int o = 0;
for (int k = 0; k <= orders[2]; k++)
{
real_t kz = kv[2]->GetBotella(ijk[2] + k);
if (!kv[2]->inSpan(kz, ijk[2]+orders[2]))
{
o += (orders[0] + 2)*(orders[1] + 1);
continue;
}
ip.z = kv[2]->GetRefPoint(kz, ijk[2]+orders[2]);
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 2;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]+1; i++, o++)
{
real_t kx = kv1[0]->GetBotella(ijk[0] + i);
if (!kv1[0]->inSpan(kx, ijk[0]+orders[0]+1)) { continue; }
ip.x = kv1[0]->GetRefPoint(kx, ijk[0]+orders[0]+1);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.AdjugateJacobian().Mult(x,mx);
dofs(o) = mx(0);
}
}
}
for (int k = 0; k <= orders[2]; k++)
{
real_t kz = kv[2]->GetBotella(ijk[2] + k);
if (!kv[2]->inSpan(kz, ijk[2]+orders[2]))
{
o += (orders[0] + 1)*(orders[1] + 2);
continue;
}
ip.z = kv[2]->GetRefPoint(kz, ijk[2]+orders[2]);
for (int j = 0; j <= orders[1]+1; j++)
{
real_t ky = kv1[1]->GetBotella(ijk[1] + j);
if (!kv1[1]->inSpan(ky, ijk[1]+orders[1]+1))
{
o += orders[0] + 1;
continue;
}
ip.y = kv1[1]->GetRefPoint(ky, ijk[1]+orders[1]+1);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.AdjugateJacobian().Mult(x,mx);
dofs(o) = mx(1);
}
}
}
for (int k = 0; k <= orders[2]+1; k++)
{
real_t kz = kv1[2]->GetBotella(ijk[2] + k);
if (!kv1[2]->inSpan(kz, ijk[2]+orders[2]+1))
{
o += (orders[0] + 1)*(orders[1] + 1);
continue;
}
ip.z = kv1[2]->GetRefPoint(kz, ijk[2]+orders[2]+1);
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 1;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.AdjugateJacobian().Mult(x,mx);
dofs(o) = mx(2);
}
}
}
}
NURBS_HDiv3DFiniteElement::~NURBS_HDiv3DFiniteElement()
{
if (kv1[0]) { delete kv1[0]; }
@@ -1163,68 +817,13 @@ void NURBS_HCurl2DFiniteElement::CalcCurlShape(const IntegrationPoint &ip,
}
}
void NURBS_HCurl2DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == dof, "");
MFEM_ASSERT(vc.GetVDim() == 2, "");
Vector x(2), xm(2);
IntegrationPoint ip;
int i, j, o;
for (o = 0, j = 0; j <= orders[1]+1; j++)
{
real_t ky = kv1[1]->GetBotella(ijk[1] + j);
if (!kv1[1]->inSpan(ky, ijk[1]+orders[1]+1))
{
o += orders[0] + 1;
continue;
}
ip.y = kv1[1]->GetRefPoint(ky, ijk[1]+orders[1]+1);
for (i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.Jacobian().MultTranspose(x,xm);
dofs(o) = xm(0);
}
}
for (j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 2;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (i = 0; i <= orders[0]+1; i++, o++)
{
real_t kx = kv1[0]->GetBotella(ijk[0] + i);
if (!kv1[0]->inSpan(kx, ijk[0]+orders[0]+1)) { continue; }
ip.x = kv1[0]->GetRefPoint(kx, ijk[0]+orders[0]+1);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.Jacobian().MultTranspose(x,xm);
dofs(o) = xm(1);
}
}
}
NURBS_HCurl2DFiniteElement::~NURBS_HCurl2DFiniteElement()
{
if (kv1[0]) { delete kv1[0]; }
if (kv1[1]) { delete kv1[1]; }
}
void NURBS_HCurl3DFiniteElement::SetOrder() const
{
orders[0] = kv[0]->GetOrder();
@@ -1404,124 +1003,11 @@ void NURBS_HCurl3DFiniteElement::CalcCurlShape(const IntegrationPoint &ip,
curl_shape(o,0) = shape1_x(i)*dsy1_sz;
curl_shape(o,1) = -dshape1_x(i)*sy1_sz;
curl_shape(o,2) = 0.0;
}
}
}
}
void NURBS_HCurl3DFiniteElement::Project(VectorCoefficient &vc,
ElementTransformation &Trans,
Vector &dofs) const
{
MFEM_ASSERT(dofs.Size() == dof, "");
MFEM_ASSERT(vc.GetVDim() == 3, "");
Vector x(3), xm(3);
IntegrationPoint ip;
int o = 0;
for (int k = 0; k <= orders[2]+1; k++)
{
real_t kz = kv1[2]->GetBotella(ijk[2] + k);
if (!kv1[2]->inSpan(kz, ijk[2]+orders[2]+1))
{
o += (orders[0] + 1)*(orders[1] + 2);
continue;
}
ip.z = kv1[2]->GetRefPoint(kz, ijk[2]+orders[2]+1);
for (int j = 0; j <= orders[1]+1; j++)
{
real_t ky = kv1[1]->GetBotella(ijk[1] + j);
if (!kv1[1]->inSpan(ky, ijk[1]+orders[1]+1))
{
o += orders[0] + 1;
continue;
}
ip.y = kv1[1]->GetRefPoint(ky, ijk[1]+orders[1]+1);
for (int i = 0; i <= orders[0]; i++, o++)
{
real_t kx = kv[0]->GetBotella(ijk[0] + i);
if (!kv[0]->inSpan(kx, ijk[0]+orders[0])) { continue; }
ip.x = kv[0]->GetRefPoint(kx, ijk[0]+orders[0]);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.Jacobian().MultTranspose(x,xm);
dofs(o) = xm(0);
}
}
}
for (int k = 0; k <= orders[2]+1; k++)
{
real_t kz = kv1[2]->GetBotella(ijk[2] + k);
if (!kv1[2]->inSpan(kz, ijk[2]+orders[2]+1))
{
o += (orders[0] + 2)*(orders[1] + 1);
continue;
}
ip.z = kv1[2]->GetRefPoint(kz, ijk[2]+orders[2]+1);
for (int j = 0; j <= orders[1]; j++)
{
real_t ky = kv[1]->GetBotella(ijk[1] + j);
if (!kv[1]->inSpan(ky, ijk[1]+orders[1]))
{
o += orders[0] + 2;
continue;
}
ip.y = kv[1]->GetRefPoint(ky, ijk[1]+orders[1]);
for (int i = 0; i <= orders[0]+1; i++, o++)
{
real_t kx = kv1[0]->GetBotella(ijk[0] + i);
if (!kv1[0]->inSpan(kx, ijk[0]+orders[0]+1)) { continue; }
ip.x = kv1[0]->GetRefPoint(kx, ijk[0]+orders[0]+1);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.Jacobian().MultTranspose(x,xm);
dofs(o) = xm(1);
}
}
}
for (int k = 0; k <= orders[2]; k++)
{
real_t kz = kv[2]->GetBotella(ijk[2] + k);
if (!kv[2]->inSpan(kz, ijk[2]+orders[2]))
{
o += (orders[0] + 2)*(orders[1] + 2);
continue;
}
ip.z = kv[2]->GetRefPoint(kz, ijk[2]+orders[2]);
for (int j = 0; j <= orders[1]+1; j++)
{
real_t ky = kv1[1]->GetBotella(ijk[1] + j);
if (!kv1[1]->inSpan(ky, ijk[1]+orders[1]+1))
{
o += orders[0] + 2;
continue;
}
ip.y = kv1[1]->GetRefPoint(ky, ijk[1]+orders[1]+1);
for (int i = 0; i <= orders[0]+1; i++, o++)
{
real_t kx = kv1[0]->GetBotella(ijk[0] + i);
if (!kv1[0]->inSpan(kx, ijk[0]+orders[0]+1)) { continue; }
ip.x = kv1[0]->GetRefPoint(kx, ijk[0]+orders[0]+1);
Trans.SetIntPoint(&ip);
vc.Eval(x, Trans, ip);
Trans.Jacobian().MultTranspose(x,xm);
dofs(o) = xm(2);
}
}
}
}
NURBS_HCurl3DFiniteElement::~NURBS_HCurl3DFiniteElement()
{
if (kv1[0]) { delete kv1[0]; }
-64
View File
@@ -86,18 +86,6 @@ public:
DenseMatrix &dshape) const override;
void CalcHessian (const IntegrationPoint &ip,
DenseMatrix &hessian) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(Coefficient &coeff,
ElementTransformation &Trans, Vector &dofs) const override;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
};
/// An arbitrary order 2D NURBS element on a square
@@ -133,18 +121,6 @@ public:
DenseMatrix &dshape) const override;
void CalcHessian (const IntegrationPoint &ip,
DenseMatrix &hessian) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(Coefficient &coeff,
ElementTransformation &Trans, Vector &dofs) const override;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
};
/// An arbitrary order 3D NURBS element on a cube
@@ -185,18 +161,6 @@ public:
DenseMatrix &dshape) const override;
void CalcHessian (const IntegrationPoint &ip,
DenseMatrix &hessian) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(Coefficient &coeff,
ElementTransformation &Trans, Vector &dofs) const override;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
};
@@ -278,13 +242,6 @@ public:
void CalcDivShape(const IntegrationPoint &ip,
Vector &divshape) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
~NURBS_HDiv2DFiniteElement();
};
@@ -379,13 +336,6 @@ public:
void CalcDivShape(const IntegrationPoint &ip,
Vector &divshape) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
~NURBS_HDiv3DFiniteElement();
};
@@ -465,13 +415,6 @@ public:
void CalcCurlShape(const IntegrationPoint &ip,
DenseMatrix &curl_shape) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
~NURBS_HCurl2DFiniteElement();
};
@@ -563,13 +506,6 @@ public:
void CalcCurlShape(const IntegrationPoint &ip,
DenseMatrix &curl_shape) const override;
using FiniteElement::Project;
/** Evaluate the dofs that are defined on this element.
Dofs that can not be evaluated will remain unmodified. */
void Project(VectorCoefficient &vcoeff,
ElementTransformation &Trans, Vector &dofs) const override;
~NURBS_HCurl3DFiniteElement();
};
-302
View File
@@ -557,101 +557,6 @@ H1Pos_TriangleElement::H1Pos_TriangleElement(const int p)
}
}
const DofToQuad &H1Pos_TriangleElement::GetRaggedTensorDofToQuad(
const FiniteElement &fe, const IntegrationRule &ir,
DofToQuad::Mode mode,
Array<DofToQuad*> &dof2quad_array)
{
DofToQuad *d2q = nullptr;
MFEM_VERIFY(mode == DofToQuad::RAGGED_TENSOR, "invalid mode requested");
#if defined(MFEM_THREAD_SAFE) && defined(MFEM_USE_OPENMP)
#pragma omp critical (DofToQuad)
#endif
{
for (int i = 0; i < dof2quad_array.Size(); i++)
{
d2q = dof2quad_array[i];
if (d2q->IntRule != &ir || d2q->mode != mode) { d2q = nullptr; }
}
if (!d2q)
{
d2q = new RaggedDofToQuad;
const int ndof = fe.GetOrder() + 1; // verify
const int nqpt = (int)floor(pow(ir.GetNPoints(), 1.0/fe.GetDim()) + 0.5);
d2q->FE = &fe;
d2q->IntRule = &ir;
d2q->mode = mode;
d2q->ndof = ndof;
d2q->nqpt = nqpt;
RaggedDofToQuad *rd2q = static_cast<RaggedDofToQuad*>(d2q);
rd2q->Ba1.SetSize(nqpt*ndof);
// second component of ragged tensor basis, technically dof*(dof-1)/2 entries
rd2q->Ba2.SetSize((int)nqpt*ndof*ndof);
rd2q->Ba1t.SetSize(nqpt*ndof);
rd2q->Ba2t.SetSize((int)nqpt*ndof*ndof);
// stores first component of ragged tensor basis with order p-1, for gradients only
rd2q->Ga1.SetSize(nqpt*(ndof -1));
// stores second component of ragged tensor basis with order p-1
rd2q->Ga2.SetSize(nqpt*(ndof-1)*(ndof -1));
rd2q->Ga1t.SetSize(nqpt*(ndof -1));
rd2q->Ga2t.SetSize(nqpt*(ndof-1)*(ndof -1));
rd2q->lex_map.SetSize(ndof * ndof);
Vector shape_a1(ndof), shape_a2(ndof * ndof);
Vector shape_Ga1(ndof-1), shape_Ga2((ndof-1) * (ndof-1));
for (int i = 0; i < nqpt; i++)
{
// The first 'nqpt' points in the first dimension 'ir' have the same x-coordinates as those
// of the 1D rule (ie. (2,0) Gauss-Jacobi rule). The first 'nqpt' points in the second dimension
// 'ir' have the same y-coordinates as those of the 1D rule for second dimension (i.e. (1,0)
// Gauss-Jacobi rule). Additionally, the Bernstein PA algorithms expect evaluation of the
// component 1D bases at the Stroud nodes pulled back to the unit square, so perform the pullback
// on the fly.
const real_t x = ir.IntPoint(i).x;
const real_t y = ir.IntPoint(nqpt*i).y / (1.0 - ir.IntPoint(nqpt*i).x);
Poly_1D::CalcBernstein(ndof-1, x, shape_a1);
Poly_1D::CalcBernstein(ndof-2, x, shape_Ga1);
for (int j = 0; j < ndof; j++)
{
rd2q->Ba1t[i+nqpt*j] = rd2q->Ba1[j+ndof*i] = shape_a1(j);
if (j < ndof-1)
{
rd2q->Ga1t[i+nqpt*j] = rd2q->Ga1[j+(ndof-1)*i] = shape_Ga1(j);
Poly_1D::CalcBernstein(ndof-2-j, y, shape_Ga2);
}
Poly_1D::CalcBernstein(ndof-1-j, y, shape_a2);
for (int k = 0; k < ndof-j; k++)
{
rd2q->Ba2t[i + nqpt*(j + ndof*k)] = rd2q->Ba2[k + ndof*(j + ndof*i)] = shape_a2(
k);
if (j < ndof-1 && k < ndof-j-1)
{
rd2q->Ga2t[i + nqpt*(j + (ndof-1)*k)] = rd2q->Ga2[k + (ndof-1)*(j +
(ndof-1)*i)] = shape_Ga2(k);
}
}
}
}
// stores the mapping from 2D Bernstein multi-index (i,j,p-i-j) to the
// lexicographic DOF ordering
for (int i = 0; i < ndof; i++)
{
for (int j = 0; j < ndof-i; j++)
{
int idx = ((2 * (ndof-1) + 3) - j) * j / 2 + i;
rd2q->lex_map[j + ndof*i] = idx;
}
}
dof2quad_array.Append(d2q);
}
}
return *d2q;
}
// static method
void H1Pos_TriangleElement::CalcShape(
const int p, const real_t l1, const real_t l2, real_t *shape)
@@ -844,213 +749,6 @@ H1Pos_TetrahedronElement::H1Pos_TetrahedronElement(const int p)
}
}
const DofToQuad &H1Pos_TetrahedronElement::GetRaggedTensorDofToQuad(
const FiniteElement &fe, const IntegrationRule &ir,
DofToQuad::Mode mode,
Array<DofToQuad*> &dof2quad_array)
{
DofToQuad *d2q = nullptr;
MFEM_VERIFY(mode == DofToQuad::RAGGED_TENSOR, "invalid mode requested");
#if defined(MFEM_THREAD_SAFE) && defined(MFEM_USE_OPENMP)
#pragma omp critical (DofToQuad)
#endif
{
for (int i = 0; i < dof2quad_array.Size(); i++)
{
d2q = dof2quad_array[i];
if (d2q->IntRule != &ir || d2q->mode != mode) { d2q = nullptr; }
}
if (!d2q)
{
d2q = new RaggedDofToQuad;
const int ndof = fe.GetOrder() + 1; // verify
const int nqpt = (int)floor(pow(ir.GetNPoints(), 1.0/fe.GetDim()) + 0.5);
const int basis_dim2d = ndof*(ndof+1) / 2;
const int basis_dim3d = ndof*(ndof+1)*(ndof+2) / 6;
const int basis_dim2d_diff = (ndof-1)*(ndof) / 2;
const int basis_dim3d_diff = (ndof-1)*(ndof)*(ndof+1) / 6;
d2q->FE = &fe;
d2q->IntRule = &ir;
d2q->mode = mode;
d2q->ndof = ndof;
d2q->nqpt = nqpt;
RaggedDofToQuad *rd2q = static_cast<RaggedDofToQuad*>(d2q);
rd2q->Ba1.SetSize(nqpt * ndof);
// second component of ragged tensor basis, technically dof*(dof-1)/2 entries
rd2q->Ba2.SetSize(nqpt * basis_dim2d);
// third component of ragged tensor basis, technically dof*(dof-1)/2 entries
rd2q->Ba3.SetSize(nqpt * basis_dim3d);
rd2q->Ba1t.SetSize(nqpt * ndof);
rd2q->Ba2t.SetSize(nqpt * basis_dim2d);
rd2q->Ba3t.SetSize(nqpt * basis_dim3d);
// stores first component of ragged tensor basis with order p-1, for gradients only
rd2q->Ga1.SetSize(nqpt * (ndof-1));
// stores second component of ragged tensor basis with order p-1
rd2q->Ga2.SetSize(nqpt * basis_dim2d_diff);
// stores third component of ragged tensor basis with order p-1
rd2q->Ga3.SetSize(nqpt * basis_dim3d_diff);
rd2q->Ga1t.SetSize(nqpt * (ndof-1));
rd2q->Ga2t.SetSize(nqpt * basis_dim2d_diff);
rd2q->Ga3t.SetSize(nqpt * basis_dim3d_diff);
rd2q->lex_map.SetSize(ndof * ndof * ndof);
rd2q->forward_map2d_diff.SetSize((ndof-1) * (ndof-1));
rd2q->forward_map3d_diff.SetSize((ndof-1) * (ndof-1) * (ndof-1));
rd2q->inverse_map2d_diff.SetSize(2 * basis_dim2d_diff);
rd2q->inverse_map3d_diff.SetSize(3 * basis_dim3d_diff);
rd2q->forward_map2d_mass.SetSize(ndof * ndof);
rd2q->forward_map3d_mass.SetSize(ndof * ndof * ndof);
rd2q->inverse_map2d_mass.SetSize(2 * basis_dim2d);
rd2q->inverse_map3d_mass.SetSize(2 * basis_dim3d);
// forward and inverse maps for multi-index to collpased 1d index for diffusion, can combine
// these four loops, but need four idx's and clause for shorter diff loops
int idx = 0;
for (int i = 0; i < ndof-1; i++)
{
for (int j = 0; j < ndof-i-1; j++)
{
rd2q->forward_map2d_diff[j + (ndof-1)*i] = idx;
rd2q->inverse_map2d_diff[2*idx] = i;
rd2q->inverse_map2d_diff[1 + 2*idx] = j;
idx++;
}
}
idx = 0;
for (int k = 0; k < ndof-1; k++)
{
for (int j = 0; j < ndof-k-1; j++)
{
for (int i = 0; i < ndof-k-j-1; i++)
{
rd2q->forward_map3d_diff[k + (ndof-1)*(j + (ndof-1)*i)] = idx;
rd2q->inverse_map3d_diff[3*idx] = i;
rd2q->inverse_map3d_diff[1 + 3*idx] = j;
rd2q->inverse_map3d_diff[2 + 3*idx] = k;
idx++;
}
}
}
// forward and inverse maps for multi-index to collpased 1d index for mass
idx = 0;
for (int j = 0; j < ndof; j++)
{
for (int i = 0; i < ndof-j; i++)
{
rd2q->forward_map2d_mass[j + ndof*i] = idx;
rd2q->inverse_map2d_mass[2*idx] = i;
rd2q->inverse_map2d_mass[1 + 2*idx] = j;
idx++;
}
}
idx = 0;
for (int k = 0; k < ndof; k++)
{
for (int j = 0; j < ndof-k; j++)
{
for (int i = 0; i < ndof-k-j; i++)
{
rd2q->forward_map3d_mass[k + ndof*(j + ndof*i)] = idx;
rd2q->inverse_map3d_mass[2*idx] = i;
rd2q->inverse_map3d_mass[1 + 2*idx] = j;
// d2q->inverse_map3d_mass[2 + 3*idx] = k;
idx++;
}
}
}
Vector shape_a1(ndof), shape_a2(ndof * ndof), shape_a3(ndof * ndof * ndof);
Vector shape_Ga1(ndof-1), shape_Ga2(ndof-1), shape_Ga3(ndof-1);
for (int i = 0; i < nqpt; i++)
{
// The first 'nqpt' points in the first dimension 'ir' have the same x-coordinates as those
// of the 1D rule (ie. (2,0) Gauss-Jacobi rule). The first 'nqpt' points in the second dimension
// 'ir' have the same y-coordinates as those of the 1D rule for second dimension (i.e. (1,0)
// Gauss-Jacobi rule). The first 'nqpt' points in the third dimension have the same z-coordinates
// as those of the 1D rule for the third dimension (i.e. Gauss-Legendre rule). Additionally,
// the Bernstein PA algorithms expect evaluation of the component 1D bases at the Stroud nodes
// pulled back to the unit cube, so perform the pullback on the fly.
const real_t x = ir.IntPoint(i).x;
const real_t y = ir.IntPoint(nqpt*i).y / (1.0 - ir.IntPoint(nqpt*i).x);
const real_t z = ir.IntPoint(nqpt*nqpt*i).z / (1.0 - ir.IntPoint(
nqpt*nqpt*i).x - ir.IntPoint(nqpt*nqpt*i).y);
Poly_1D::CalcBernstein(ndof-1, x, shape_a1);
Poly_1D::CalcBernstein(ndof-2, x, shape_Ga1);
for (int j = 0; j < ndof; j++)
{
rd2q->Ba1t[i+nqpt*j] = rd2q->Ba1[j+ndof*i] = shape_a1(j);
if (j < ndof-1)
{
rd2q->Ga1t[i+nqpt*j] = rd2q->Ga1[j+(ndof-1)*i] = shape_Ga1(j);
Poly_1D::CalcBernstein(ndof-2-j, y, shape_Ga2);
}
Poly_1D::CalcBernstein(ndof-1-j, y, shape_a2);
for (int k = 0; k < ndof-j; k++)
{
const int a_2d_mass = rd2q->forward_map2d_mass[k + ndof*j];
rd2q->Ba2t[i + nqpt*a_2d_mass] = rd2q->Ba2[a_2d_mass + basis_dim2d*i] =
shape_a2(
k);
if (j < ndof-1 && k < ndof-j-1)
{
const int a_2d_diff = rd2q->forward_map2d_diff[k + (ndof-1)*j];
rd2q->Ga2t[i + nqpt*a_2d_diff] = rd2q->Ga2[a_2d_diff + basis_dim2d_diff*i] =
shape_Ga2(k);
Poly_1D::CalcBernstein(ndof-2-j-k, z, shape_Ga3);
}
Poly_1D::CalcBernstein(ndof-1-j-k, z, shape_a3);
for (int m = 0; m < ndof-j-k; m++)
{
const int a_3d_mass = rd2q->forward_map3d_mass[m + ndof*(k + ndof*j)];
rd2q->Ba3t[i + nqpt*a_3d_mass] = rd2q->Ba3[a_3d_mass + basis_dim3d*i] =
shape_a3(
m);
if (j < ndof-1 && k < ndof-j-1 && m < ndof-j-k-1)
{
// // collapsed 1D access
// d2q->Ga3[i + nqpt*(m + d2q->offset3d[k + (ndof-1)*j])] = shape_Ga3(m);
// collapsed 1D access with forward mapping
const int a_3d_diff = rd2q->forward_map3d_diff[m + (ndof-1)*(k + (ndof-1)*j)];
rd2q->Ga3t[i + nqpt*a_3d_diff] = rd2q->Ga3[a_3d_diff + basis_dim3d_diff*i] =
shape_Ga3(m);
}
}
}
}
}
// stores the mapping from 3D Bernstein multi-index (i,j,k,p-i-j-k) to the
// lexicographic DOF ordering
int p = ndof - 1;
for (int i = 0; i < ndof; i++)
{
for (int j = 0; j < ndof-i; j++)
{
for (int k = 0; k < ndof-i-j; k++)
{
int dof = (p+1)*(p+2)*(p+3) / 6;
int tet = (p-k)*(p-k+1)*(p-k+2) / 6;
int tri = (p+1-k-j)*(p+2-k-j)/2;
int multi_idx = dof - tet - tri + i;
rd2q->lex_map[k + ndof*(j + ndof*i)] = multi_idx;
}
}
}
dof2quad_array.Append(d2q);
}
}
return *d2q;
}
// static method
void H1Pos_TetrahedronElement::CalcShape(
const int p, const real_t l1, const real_t l2, const real_t l3,
-30
View File
@@ -191,21 +191,6 @@ public:
/// Construct the H1Pos_TriangleElement of order @a p
H1Pos_TriangleElement(const int p);
const DofToQuad &GetDofToQuad(const IntegrationRule &ir,
DofToQuad::Mode mode) const override
{
return (mode == DofToQuad::RAGGED_TENSOR) ?
GetRaggedTensorDofToQuad(*this, ir, mode, dof2quad_array) :
FiniteElement::GetDofToQuad(ir, mode);
}
static const DofToQuad &GetRaggedTensorDofToQuad(
const FiniteElement &fe, const IntegrationRule &ir,
DofToQuad::Mode mode,
Array<DofToQuad*> &dof2quad_array);
const Array<int> &GetDofMap() const { return dof_map; }
// The size of shape is (p+1)(p+2)/2 (dof).
static void CalcShape(const int p, const real_t x, const real_t y,
real_t *shape);
@@ -235,21 +220,6 @@ public:
/// Construct the H1Pos_TetrahedronElement of order @a p
H1Pos_TetrahedronElement(const int p);
const DofToQuad &GetDofToQuad(const IntegrationRule &ir,
DofToQuad::Mode mode) const override
{
return (mode == DofToQuad::RAGGED_TENSOR) ?
GetRaggedTensorDofToQuad(*this, ir, mode, dof2quad_array) :
FiniteElement::GetDofToQuad(ir, mode);
}
static const DofToQuad &GetRaggedTensorDofToQuad(
const FiniteElement &fe, const IntegrationRule &ir,
DofToQuad::Mode mode,
Array<DofToQuad*> &dof2quad_array);
const Array<int> &GetDofMap() const { return dof_map; }
// The size of shape is (p+1)(p+2)(p+3)/6 (dof).
static void CalcShape(const int p, const real_t x, const real_t y,
const real_t z, real_t *shape);
+3 -3
View File
@@ -2006,7 +2006,7 @@ RT_R1D_SegmentElement::RT_R1D_SegmentElement(const int p,
const int cb_type,
const int ob_type)
: VectorFiniteElement(1, Geometry::SEGMENT, 3 * p + 4, p + 1,
H_DIV_R1D, FunctionSpace::Pk),
H_DIV, FunctionSpace::Pk),
dof2nk(dof),
cbasis1d(poly1d.GetBasis(p + 1, VerifyClosed(cb_type))),
obasis1d(poly1d.GetBasis(p, VerifyOpen(ob_type)))
@@ -2281,7 +2281,7 @@ const real_t RT_R2D_SegmentElement::nk[2] = { 0.,1.};
RT_R2D_SegmentElement::RT_R2D_SegmentElement(const int p,
const int ob_type)
: VectorFiniteElement(1, Geometry::SEGMENT, p + 1, p + 1,
H_DIV_R2D, FunctionSpace::Pk),
H_DIV, FunctionSpace::Pk),
dof2nk(dof),
obasis1d(poly1d.GetBasis(p, VerifyOpen(ob_type)))
{
@@ -2392,7 +2392,7 @@ void RT_R2D_SegmentElement::LocalInterpolation(const VectorFiniteElement &cfe,
RT_R2D_FiniteElement::RT_R2D_FiniteElement(int p, Geometry::Type G, int Do,
const real_t *nk_fe)
: VectorFiniteElement(2, G, Do, p + 1,
H_DIV_R2D, FunctionSpace::Pk),
H_DIV, FunctionSpace::Pk),
nk(nk_fe),
dof_map(dof),
dof2nk(dof)
-6
View File
@@ -510,9 +510,6 @@ public:
RT_R2D_SegmentElement(const int p,
const int ob_type = BasisType::GaussLegendre);
int GetPhysRangeDim(int space_dim) const override { return 2; }
int GetPhysCurlDim(int space_dim) const override { return 0; }
void CalcVShape(const IntegrationPoint &ip,
DenseMatrix &shape) const override;
@@ -550,9 +547,6 @@ private:
DenseMatrix &I) const;
public:
int GetPhysRangeDim(int space_dim) const override { return 3; }
int GetPhysCurlDim(int space_dim) const override { return 0; }
using FiniteElement::CalcVShape;
void CalcVShape(ElementTransformation &Trans,
+13 -40
View File
@@ -111,36 +111,36 @@ public:
| :------: | :---: | :---: | :-------: | :-----: | :---: |
| H1_[DIM]_[ORDER] | H1 | * | 1 | VALUE | H1 nodal elements |
| H1@[BTYPE]_[DIM]_[ORDER] | H1 | * | * | VALUE | H1 nodal elements |
| H1Pos_[DIM]_[ORDER] | H1 | * | 2 | VALUE | H1 nodal elements |
| H1Pos_[DIM]_[ORDER] | H1 | * | 1 | VALUE | H1 nodal elements |
| H1Pos_Trace_[DIM]_[ORDER] | H^{1/2} | * | 2 | VALUE | H^{1/2}-conforming trace elements for H1 defined on the interface between mesh elements (faces,edges,vertices) |
| H1_Trace_[DIM]_[ORDER] | H^{1/2} | * | 1 | VALUE | H^{1/2}-conforming trace elements for H1 defined on the interface between mesh elements (faces,edges,vertices) |
| H1_Trace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | * | VALUE | H^{1/2}-conforming trace elements for H1 defined on the interface between mesh elements (faces,edges,vertices) |
| H1_Trace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | 1 | VALUE | H^{1/2}-conforming trace elements for H1 defined on the interface between mesh elements (faces,edges,vertices) |
| ND_[DIM]_[ORDER] | H(curl) | * | 1 / 0 | H_CURL | Nedelec vector elements |
| ND@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(curl) | * | * / * | H_CURL | Nedelec vector elements |
| ND_Trace_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | H_CURL | H^{1/2}-conforming trace elements for H(curl) defined on the interface between mesh elements (faces,edges) |
| ND_Trace@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H^{1/2} | * | * / * | H_CURL | H^{1/2}-conforming trace elements for H(curl) defined on the interface between mesh elements (faces,edges) |
| ND_Trace_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | H_CURL | H^{1/2}-conforming trace elements for H(curl) defined on the interface between mesh elements (faces) |
| ND_Trace@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | H_CURL | H^{1/2}-conforming trace elements for H(curl) defined on the interface between mesh elements (faces) |
| ND_R1D_[DIM]_[ORDER] | H(curl) | * | 1 / 0 | H_CURL | 3D H(curl)-conforming Nedelec vector elements in 1D. |
| ND_R1D@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(curl) | * | * / * | H_CURL | 3D H(curl)-conforming Nedelec vector elements in 1D. |
| ND_R2D_[DIM]_[ORDER] | H(curl) | * | 1 / 0 | H_CURL | 3D H(curl)-conforming Nedelec vector elements in 2D. |
| ND_R2D@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(curl) | * | * / * | H_CURL | 3D H(curl)-conforming Nedelec vector elements in 2D. |
| RT_[DIM]_[ORDER] | H(div) | * | 1 / 0 | H_DIV | Raviart-Thomas vector elements |
| RT@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(div) | * | * / * | H_DIV | Raviart-Thomas vector elements |
| RT_Trace_[DIM]_[ORDER] | H^{1/2} | * | 0 | INTEGRAL | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_ValTrace_[DIM]_[ORDER] | H^{1/2} | * | 0 | VALUE | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_Trace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | * | INTEGRAL | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_ValTrace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | * | VALUE | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_Trace_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | INTEGRAL | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_ValTrace_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | VALUE | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_Trace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | INTEGRAL | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_ValTrace@[BTYPE]_[DIM]_[ORDER] | H^{1/2} | * | 1 / 0 | VALUE | H^{1/2}-conforming trace elements for H(div) defined on the interface between mesh elements (faces) |
| RT_R1D_[DIM]_[ORDER] | H(div) | * | 1 / 0 | H_DIV | 3D H(div)-conforming Raviart-Thomas vector elements in 1D. |
| RT_R1D@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(div) | * | * / * | H_DIV | 3D H(div)-conforming Raviart-Thomas vector elements in 1D. |
| RT_R2D_[DIM]_[ORDER] | H(div) | * | 1 / 0 | H_DIV | 3D H(div)-conforming Raviart-Thomas vector elements in 2D. |
| RT_R2D@[CBTYPE][OBTYPE]_[DIM]_[ORDER] | H(div) | * | * / * | H_DIV | 3D H(div)-conforming Raviart-Thomas vector elements in 2D. |
| L2_[DIM]_[ORDER] | L2 | * | 0 | VALUE | Discontinuous L2 elements |
| L2_T[BTYPE]_[DIM]_[ORDER] | L2 | * | * | VALUE | Discontinuous L2 elements |
| L2_T[BTYPE]_[DIM]_[ORDER] | L2 | * | 0 | VALUE | Discontinuous L2 elements |
| L2Int_[DIM]_[ORDER] | L2 | * | 0 | INTEGRAL | Discontinuous L2 elements |
| L2Int_T[BTYPE]_[DIM]_[ORDER] | L2 | * | * | INTEGRAL | Discontinuous L2 elements |
| L2Int_T[BTYPE]_[DIM]_[ORDER] | L2 | * | 0 | INTEGRAL | Discontinuous L2 elements |
| DG_Iface_[DIM]_[ORDER] | - | * | 0 | VALUE | Discontinuous elements on the interface between mesh elements (faces) |
| DG_Iface@[BTYPE]_[DIM]_[ORDER] | - | * | * | VALUE | Discontinuous elements on the interface between mesh elements (faces) |
| DG_Iface@[BTYPE]_[DIM]_[ORDER] | - | * | 0 | VALUE | Discontinuous elements on the interface between mesh elements (faces) |
| DG_IntIface_[DIM]_[ORDER] | - | * | 0 | INTEGRAL | Discontinuous elements on the interface between mesh elements (faces) |
| DG_IntIface@[BTYPE]_[DIM]_[ORDER] | - | * | * | INTEGRAL | Discontinuous elements on the interface between mesh elements (faces) |
| DG_IntIface@[BTYPE]_[DIM]_[ORDER] | - | * | 0 | INTEGRAL | Discontinuous elements on the interface between mesh elements (faces) |
| NURBS[ORDER] | - | * | - | VALUE | Non-Uniform Rational B-Splines (NURBS) elements |
| LinearNonConf3D | - | 1 | 1 | VALUE | Piecewise-linear nonconforming finite elements in 3D |
| CrouzeixRaviart | - | - | - | - | Crouzeix-Raviart nonconforming elements in 2D |
@@ -172,7 +172,7 @@ public:
| :------: | :--------: |
| [DIM] | Dimension of the elements (1D, 2D, 3D) |
| [ORDER] | Approximation order of the elements (P0, P1, P2, ...) |
| [BTYPE] | BasisType of the element (0-GaussLegendre, 1-GaussLobatto, 2-Bernstein, 3-OpenUniform, 4-CloseUniform, 5-OpenHalfUniform 6-Serendipity 7-ClosedGL 8-IntegratedGLL) |
| [BTYPE] | BasisType of the element (0-GaussLegendre, 1 - GaussLobatto, 2-Bernstein, 3-OpenUniform, 4-CloseUniform, 5-OpenHalfUniform) |
| [OBTYPE] | Open BasisType of the element for elements which have both types |
| [CBTYPE] | Closed BasisType of the element for elements which have both types |
@@ -250,14 +250,6 @@ public:
its GetOrder() method. */
virtual FiniteElementCollection *Clone(int p) const;
/** @brief Return the order parameter used to construct this collection.
* This differs from GetOrder() depending on the collection type. */
virtual int GetConstructorOrder() const
{
MFEM_ABORT("Collection " << Name() << " does not support GetConstructorOrder");
return -1;
}
protected:
const int base_p; ///< Order as returned by GetOrder().
@@ -322,9 +314,6 @@ public:
FiniteElementCollection *Clone(int p) const override
{ return new H1_FECollection(p, dim, b_type); }
int GetConstructorOrder() const override
{ return base_p; }
virtual ~H1_FECollection();
};
@@ -354,10 +343,6 @@ class H1_Trace_FECollection : public H1_FECollection
public:
H1_Trace_FECollection(const int p, const int dim,
const int btype = BasisType::GaussLobatto);
FiniteElementCollection *Clone(int p) const override
{ return new H1_Trace_FECollection(p, dim+1, b_type); }
};
/// Arbitrary order "L2-conforming" discontinuous finite elements.
@@ -411,9 +396,6 @@ public:
FiniteElementCollection *Clone(int p) const override
{ return new L2_FECollection(p, dim, b_type, m_type); }
int GetConstructorOrder() const override
{ return base_p; }
virtual ~L2_FECollection();
};
@@ -474,9 +456,6 @@ public:
FiniteElementCollection *Clone(int p) const override
{ return new RT_FECollection(p, dim, cb_type, ob_type); }
int GetConstructorOrder() const override
{ return base_p-1; }
virtual ~RT_FECollection();
};
@@ -557,9 +536,6 @@ public:
FiniteElementCollection *Clone(int p) const override
{ return new ND_FECollection(p, dim, cb_type, ob_type); }
int GetConstructorOrder() const override
{ return dim>1 ? base_p : base_p+1; }
virtual ~ND_FECollection();
};
@@ -572,9 +548,6 @@ public:
ND_Trace_FECollection(const int p, const int dim,
const int cb_type = BasisType::GaussLobatto,
const int ob_type = BasisType::GaussLegendre);
FiniteElementCollection *Clone(int p) const override
{ return new ND_Trace_FECollection(p, dim+1, cb_type, ob_type); }
};
/// Arbitrary order 3D H(curl)-conforming Nedelec finite elements in 1D.
+34 -79
View File
@@ -282,7 +282,14 @@ int FiniteElementSpace::DofToVDof(int dof, int vd, int ndofs_) const
void FiniteElementSpace::AdjustVDofs(Array<int> &vdofs)
{
int n = vdofs.Size(), *vdof = vdofs;
for (int i = 0; i < n; i++) { vdof[i] = UnsignIndex(vdof[i]); }
for (int i = 0; i < n; i++)
{
int j;
if ((j = vdof[i]) < 0)
{
vdof[i] = -1-j;
}
}
}
void FiniteElementSpace::GetElementVDofs(int i, Array<int> &vdofs,
@@ -476,14 +483,13 @@ void FiniteElementSpace::ReorderElementToDofTable()
for (int k = 0, dof_counter = 0; k < nnz; k++)
{
const int sdof = J[k]; // signed dof
const int dof = UnsignIndex(sdof);
const int dof = (sdof < 0) ? -1-sdof : sdof;
int new_dof = dof_marker[dof];
if (new_dof < 0)
{
dof_marker[dof] = new_dof = dof_counter++;
}
// Preserve the sign of sdof
J[k] = (sdof < 0) ? FlipIndexSign(new_dof) : new_dof;
J[k] = (sdof < 0) ? -1-new_dof : new_dof; // preserve the sign of sdof
}
}
@@ -541,7 +547,7 @@ void MarkDofs(const Array<int> &dofs, Array<int> &mark_array)
{
for (auto d : dofs)
{
mark_array[UnsignIndex(d)] = -1;
mark_array[d >= 0 ? d : -1 - d] = -1;
}
}
@@ -925,7 +931,7 @@ void FiniteElementSpace::AddDependencies(
if (std::abs(coef) > 1e-12)
{
const int mdof = master_dofs[j];
if (mdof != sdof && mdof != FlipIndexSign(sdof))
if (mdof != sdof && mdof != (-1-sdof))
{
deps.Add(sdof, mdof, coef);
}
@@ -1018,7 +1024,7 @@ int FiniteElementSpace::GetDegenerateFaceDofs(int index, Array<int> &dofs,
// FiniteElementSpace::AddDependencies.
Array<int> edof;
int order = GetEdgeDofs(FlipIndexSign(index), edof, variant);
int order = GetEdgeDofs(-1 - index, edof, variant);
int nv = fec->DofForGeometry(Geometry::POINT);
int ne = fec->DofForGeometry(Geometry::SEGMENT);
@@ -1510,76 +1516,36 @@ const FaceRestriction *FiniteElementSpace::GetFaceRestriction(
const bool is_dg_space = IsDGSpace();
const L2FaceValues m = (is_dg_space && mul==L2FaceValues::DoubleValued) ?
L2FaceValues::DoubleValued : L2FaceValues::SingleValued;
auto key = std::make_tuple(is_dg_space, f_ordering, type, m);
key_face key = std::make_tuple(is_dg_space, f_ordering, type, m);
auto itr = L2F.find(key);
if (itr != L2F.end())
{
return itr->second.get();
return itr->second;
}
else
{
std::unique_ptr<FaceRestriction> res;
FaceRestriction *res;
if (is_dg_space)
{
if (Conforming())
{
res.reset(new L2FaceRestriction(*this, f_ordering, type, m));
res = new L2FaceRestriction(*this, f_ordering, type, m);
}
else
{
res.reset(new NCL2FaceRestriction(*this, f_ordering, type, m));
res = new NCL2FaceRestriction(*this, f_ordering, type, m);
}
}
else if (dynamic_cast<const DG_Interface_FECollection*>(fec))
{
res.reset(new L2InterfaceFaceRestriction(*this, f_ordering, type));
res = new L2InterfaceFaceRestriction(*this, f_ordering, type);
}
else
{
res.reset(new ConformingFaceRestriction(*this, f_ordering, type));
res = new ConformingFaceRestriction(*this, f_ordering, type);
}
return L2F.emplace(key, std::move(res)).first->second.get();
}
}
const InterpolationManager &FiniteElementSpace::GetInterpolationManager(
ElementDofOrdering f_ordering, FaceType type) const
{
const auto key = make_tuple(f_ordering, type);
auto it = interpolations.find(key);
if (it != interpolations.end())
{
return *it->second;
}
else
{
auto interp = make_unique<InterpolationManager>(*this, f_ordering, type);
int face_idx = 0;
for (int f = 0; f < mesh->GetNumFacesWithGhost(); ++f)
{
Mesh::FaceInformation face = mesh->GetFaceInformation(f);
if (!face.IsOfFaceType(type) || face.IsNonconformingCoarse())
{
continue;
}
if (face.IsConforming() || face.IsBoundary())
{
interp->RegisterFaceConformingInterpolation(face, face_idx);
}
else
{
interp->RegisterFaceCoarseToFineInterpolation(face, face_idx);
}
++face_idx;
}
// Transform the interpolation matrix map into contiguous memory.
interp->LinearizeInterpolatorMapIntoVector();
interp->InitializeNCInterpConfig();
return *interpolations.emplace(key, std::move(interp)).first->second;
L2F[key] = res;
return res;
}
}
@@ -1704,8 +1670,8 @@ SparseMatrix *FiniteElementSpace::RefinementMatrix_main(
for (int i = 0; i < fine_ldof; i++)
{
const int r = DofToVDof(dofs[i], vd);
const int m = UnsignIndex(r);
int r = DofToVDof(dofs[i], vd);
int m = (r >= 0) ? r : (-1 - r);
if (!mark[m])
{
@@ -1766,7 +1732,7 @@ SparseMatrix *FiniteElementSpace::VariableOrderRefinementMatrix(
for (int i = 0; i < fine_ldof; i++)
{
const int r = DofToVDof(dofs[i], vd);
const int m = UnsignIndex(r);
int m = (r >= 0) ? r : (-1 - r);
if (!mark[m])
{
@@ -2476,8 +2442,8 @@ SparseMatrix* FiniteElementSpace::DerefinementMatrix(int old_ndofs,
{
if (!std::isfinite(lR(i, 0))) { continue; }
const int r = DofToVDof(dofs[i], vd);
const int m = UnsignIndex(r);
int r = DofToVDof(dofs[i], vd);
int m = (r >= 0) ? r : (-1 - r);
if (is_dg || !mark[m])
{
@@ -3195,7 +3161,7 @@ void FiniteElementSpace::CalcEdgeFaceVarOrders(
else
{
// degenerate face (i.e., edge-face constraint)
slave_orders |= edge_orders[FlipIndexSign(slave.index)];
slave_orders |= edge_orders[-1 - slave.index];
}
}
@@ -3934,16 +3900,6 @@ const FiniteElement *FiniteElementSpace::GetBE(int i) const
return BE;
}
const FiniteElement *FiniteElementSpace::GetTypicalBE() const
{
if (mesh->GetNBE() > 0) { return GetBE(0); }
Geometry::Type geom = mesh->GetTypicalFaceGeometry();
const FiniteElement *be = fec->FiniteElementForGeometry(geom);
MFEM_VERIFY(be != nullptr, "Could not determine a typical BE!");
return be;
}
const FiniteElement *FiniteElementSpace::GetFaceElement(int i) const
{
MFEM_VERIFY(!IsVariableOrder(), "not implemented");
@@ -3974,11 +3930,6 @@ const FiniteElement *FiniteElementSpace::GetFaceElement(int i) const
return fe;
}
const FiniteElement *FiniteElementSpace::GetTypicalFaceElement() const
{
return fec->FiniteElementForGeometry(mesh->GetTypicalFaceGeometry());
}
const FiniteElement *FiniteElementSpace::GetEdgeElement(int i,
int variant) const
{
@@ -4018,8 +3969,11 @@ void FiniteElementSpace::Destroy()
delete E2Q_array[i];
}
E2Q_array.SetSize(0);
for (auto &x : L2F)
{
delete x.second;
}
L2F.clear();
interpolations.clear();
for (int i = 0; i < E2IFQ_array.Size(); i++)
{
delete E2IFQ_array[i];
@@ -4631,8 +4585,9 @@ FiniteElementCollection *FiniteElementSpace::Load(Mesh *m, std::istream &input)
ElementDofOrdering GetEVectorOrdering(const FiniteElementSpace& fes)
{
return (UsesTensorBasis(fes) || fes.UsesRaggedTensorBasis()) ?
return UsesTensorBasis(fes)?
ElementDofOrdering::LEXICOGRAPHIC:
ElementDofOrdering::NATIVE;
}
} // namespace mfem
+14 -35
View File
@@ -13,7 +13,6 @@
#define MFEM_FESPACE
#include "../config/config.hpp"
#include "../general/hash_util.hpp"
#include "../linalg/ordering.hpp"
#include "../linalg/sparsemat.hpp"
#include "../mesh/mesh.hpp"
@@ -321,11 +320,18 @@ protected:
mutable OperatorHandle L2E_nat, L2E_lex;
/// The face restriction operators, see GetFaceRestriction().
using key_face = std::tuple<bool, ElementDofOrdering, FaceType, L2FaceValues>;
mutable std::unordered_map<key_face,std::unique_ptr<FaceRestriction>,
TupleHasher> L2F;
mutable std::unordered_map<std::tuple<ElementDofOrdering,FaceType>,
std::unique_ptr<InterpolationManager>, TupleHasher> interpolations;
struct key_hash
{
std::size_t operator()(const key_face& k) const
{
return std::get<0>(k)
+ 2 * (int)std::get<1>(k)
+ 4 * (int)std::get<2>(k)
+ 8 * (int)std::get<3>(k);
}
};
using map_L2F = std::unordered_map<const key_face,FaceRestriction*,key_hash>;
mutable map_L2F L2F;
mutable Array<QuadratureInterpolator*> E2Q_array;
mutable Array<FaceQuadratureInterpolator*> E2IFQ_array;
@@ -745,9 +751,6 @@ public:
ElementDofOrdering f_ordering, FaceType,
L2FaceValues mul = L2FaceValues::DoubleValued) const;
const InterpolationManager &GetInterpolationManager(
ElementDofOrdering f_ordering, FaceType type) const;
/** @brief Return a QuadratureInterpolator that interpolates E-vectors to
quadrature point values and/or derivatives (Q-vectors). */
/** An E-vector represents the element-wise discontinuous version of the FE
@@ -839,7 +842,7 @@ public:
Note: For vector-valued elements, the results pads up the range dimension
to the spatial dimension. E.g., consider a stack of 5 vector-valued
elements each representing 2D vectors, living in a 3 dimensional space.
Then this function would give 15, not 10.
Then this fucntion would give 15, not 10.
*/
int GetVectorDim() const;
@@ -1150,7 +1153,7 @@ public:
/// Helper to return the DOF associated with a sign encoded DOF
static inline int DecodeDof(int dof)
{ return UnsignIndex(dof); }
{ return (dof >= 0) ? dof : (-1 - dof); }
/// Helper to determine the DOF and sign of a sign encoded DOF
static inline int DecodeDof(int dof, real_t& sign)
@@ -1323,24 +1326,12 @@ public:
associated with i'th boundary face in the mesh object. */
const FiniteElement *GetBE(int i) const;
/// @brief Return a "typical" boundary element.
///
/// This can be used in situations where the local mesh partition may be
/// empty.
const FiniteElement *GetTypicalBE() const;
/** @brief Returns pointer to the FiniteElement in the FiniteElementCollection
associated with i'th face in the mesh object. Faces in this case refer
to the MESHDIM-1 primitive so in 2D they are segments and in 1D they are
points.*/
const FiniteElement *GetFaceElement(int i) const;
/// @brief Return a "typical" face element.
///
/// This can be used in situations where the local mesh partition may be
/// empty.
const FiniteElement *GetTypicalFaceElement() const;
/** @brief Returns pointer to the FiniteElement in the FiniteElementCollection
associated with i'th edge in the mesh object. */
const FiniteElement *GetEdgeElement(int i, int variant = 0) const;
@@ -1514,18 +1505,6 @@ public:
return dynamic_cast<const L2_FECollection*>(fec) != NULL;
}
/// @brief Return true if the mesh contains only one topology, the elements are
/// all triangles or tetrahedrons, and the elements are ragged tensor elements
/// i.e. Bernstein/positive basis.
bool UsesRaggedTensorBasis() const
{
bool simplex = this->GetMesh()->IsSimplexMesh();
bool positive =
dynamic_cast<const mfem::H1Pos_TriangleElement *>(this->GetTypicalFE()) ||
dynamic_cast<const mfem::H1Pos_TetrahedronElement *>(this->GetTypicalFE());
return simplex && positive;
}
/** In variable-order spaces on nonconforming (NC) meshes, this function
controls whether strict conformity is enforced in cases where coarse
edges/faces have higher polynomial order than their fine NC neighbors.
+134 -1331
View File
File diff suppressed because it is too large Load Diff

Some files were not shown because too many files have changed in this diff Show More