Reorganize material models with Eigen #318

aabrown100-git · 2024-12-18T19:24:13Z

Current situation

Resolves #297

Release Notes

Removes struct implementation using C-arrays (struct_3d_carray() and mat_models_carray.h)
Consolidates get_pk2cc() (for struct) and get_pk2cc_dev() (ustruct)
Consolidates some common tensor operations in get_pk2cc() into functions
Implements get_pk2cc() with fixed-size Eigen matrices and tensors

Documentation

Testing

No new tests, but all integration and unit tests pass
Updated tests/cases/struct/LV_HolzapfelOgdenModifiedAnisotropy_passive/result_001.vtu, since original get_pk2cc() implementation was missing pressure component; see commit d1cf925

Code of Conduct & Contributing Guidelines

I agree to follow the Code of Conduct and Contributing Guidelines.

…update HO-ma reference solutions, remove mat_models_carray.h includes

…to header; using fixed size Matrix and Tensors, instead of just Tensors. Compiles and passes integration tests/unit tests that use NeoHookean

…to S_iso and CC_iso; Add several required tensor operations for Eigen Matrices/Tensors

…auto instead of EigenTensor<nsd> for PP. 2) fix typo for ten_symm_prod_eigen()

…tation, only modifying to use Eigen tensor functions.

…nded for Eigen

…faster for some reason

…into reorg_mat_models

…ultiPhysics

…ization

…erformance

Code/Source/solver/mat_fun.h

Code/Source/solver/mat_models.cpp

…se Arrays instead of C-arrays.

aabrown100-git · 2024-12-19T00:38:57Z

Very nice work, @aabrown100-git!

I saw that there is a mat_models_Dave_array.cpp, which is marked as a new file. Is that just incorrectly tracked by git, or is that really a new file? Do we still want to redundant implementations, Eigen vs. Dave array?

Oops sorry! mat_models_Dave_array.cpp was just a temporary file I created while converting to Eigen. I just removed it!

Code/Source/solver/mat_models.cpp

ktbolt

@aabrown100-git Lots of good work here!

CMakeLists.txt

Code/Source/solver/mat_fun.h

Code/Source/solver/mat_models.cpp

Code/Source/solver/sv_struct.cpp

tests/unitTests/test.h

…er flags to try to fix Ubuntu timeout issue

aabrown100-git · 2024-12-19T16:52:34Z

Looks like -march=native causes an issue on GitHub Ubuntu machines; the integration tests timed out when I included -march=native in CMakeLists.txt, but passed in the most recent commit when I removed -march=native. From the timing tests I performed, there is some improvement with -march=native, but it's not huge. Also, here's a relevant discussion why -march=native is not enabled by default. For these reasons, I'll remove -march=native from CMakeLists.txt. Perhaps we can revisit this once @zasexton completes the Eigen backend for Arrays and learns more about Eigen performance.

…pressions. Avoid overloaded function and improve readability

…nd EigenTensor aliases to Matrix and Tensor. Consolidate double_dot_product functions for tensors. Rename get_() functions to compute_(). Fix throw runtime_error bug for invalid material models.

ktbolt

It all looks good to me!

aabrown100-git · 2024-12-28T18:06:26Z

Awesome, I will merge now. Thanks @ktbolt and @mrp089!

aabrown100-git and others added 29 commits November 4, 2024 20:51

Consolidate get_pk2cc and get_pk2cc_dev(). Use only mat_models.cpp

5695bc8

Update material unit tests

8b9097d

Add voigt_to_cc, fix undefined Ja bug, fix no pressure in HO-ma bug, …

d1cf925

…update HO-ma reference solutions, remove mat_models_carray.h includes

Fix Ja calculation bug leading to ustruct failures, remove get_pk2cc_dev

bf4e282

Fix indexing bug for struct viscosity tangent terms

6fcf1ac

Remove mat_models_carray.h and commented code

efcfd7a

trying out eigen conversion

dccdc64

Change from Matrix to Tensor

30cb13e

Adding/moving templated Eigen matrix/tensor function implementations …

3a3d284

…to header; using fixed size Matrix and Tensors, instead of just Tensors. Compiles and passes integration tests/unit tests that use NeoHookean

Update eigen includes path

763b92a

Consolidate volumetric stress and elasticity contributions

2744b17

General function used by multiple models to project S_bar and CC_bar …

a57187c

…to S_iso and CC_iso; Add several required tensor operations for Eigen Matrices/Tensors

Bug fixes. 1) For some reason, code runs but is incorrect when using …

86abc97

…auto instead of EigenTensor<nsd> for PP. 2) fix typo for ten_symm_prod_eigen()

Replace some mat_fun loops with built in Eigen operations

51bf817

Add Eigen implementation for Mooney-Rivlin model

036ae7d

Adding Linear, StVK, and modified StVK. Compared to original implemen…

df75bec

…tation, only modifying to use Eigen tensor functions.

Add HGO model. Passes ustruct/tensile_adventitia_HGO

9b9806f

Add Guccione model Eigen implementation. Passes Guccione tests

74bcf58

Add Holzapfel-Ogden model. Passes HO tests

402f298

Add HO-ma model. Passes all struct/ustruct tests

dfe941a

Replacing auto with EigenMatrix<nsd>, since using auto is not recomme…

3364a32

…nded for Eigen

Replacing Eigen operations with for loops for some tensor functions, …

44bdc58

…faster for some reason

Merge branch 'main' of https://github.com/SimVascular/svMultiPhysics …

d830316

…into reorg_mat_models

Update .gitignore genBC files to reflect renaming of svFSIplus to svM…

cff0979

…ultiPhysics

Merge branch 'SimVascular:main' into reorg_mat_models

c3d6f02

Merge branch 'SimVascular:main' into reorg_mat_models

4776b6f

Add -march=native flag to CMAKE_C_FLAGS and CMAKE_CXX_FLAGS for optim…

00e95b3

…ization

Update CMakeLists.txt to clarify -march=native flag usage for Eigen p…

85f6504

…erformance

Merge branch 'SimVascular:main' into reorg_mat_models

c1fd61e

aabrown100-git requested a review from mrp089 December 18, 2024 19:27

ktbolt reviewed Dec 18, 2024

View reviewed changes

Code/Source/solver/mat_fun.h Outdated Show resolved Hide resolved

ktbolt reviewed Dec 18, 2024

View reviewed changes

Code/Source/solver/mat_fun.h Outdated Show resolved Hide resolved

ktbolt reviewed Dec 18, 2024

View reviewed changes

Code/Source/solver/mat_fun.h Outdated Show resolved Hide resolved