initializing plastCo from mu leads to SeisSol crashing

[Thomas-Ulrich]

Initializing the plastic cohesion with:

!Switch
[rho, mu, lambda]:  !Include Aqaba_rhomulambda.yaml
[bulkFriction]: !ConstantMap
  map:
    bulkFriction: 0.6
[plastCo]: !IdentityMap
  components:
    - !SphericalDomainFilter
      radius: 1800
      centre:
        x: -1540
        y: -22115
        z: -7200
      components: !ConstantMap
        map:
          plastCo: 1.0e20
    - !EvalModel
        parameters: [mu]
        model: !Switch
            [mu]: !Include Aqaba_rhomulambda.yaml
        components: !FunctionMap
           map:
              plastCo: return 0.0001*mu;
[s_xx, s_yy, s_zz, s_xy, s_yz, s_xz]: !Include Aqaba_initial_stress.yaml

Where Aqaba_rhomulambda.yaml is:

!LayeredModel
  map: !AffineMap
    matrix:
      z1: [0.0, 0.0, 1.0]
    translation:
      z1: 0
  interpolation: lower
  parameters: [rho, mu, lambda]
  nodes:
    10000.000000: [2348.000000, 1.865899e+10, 2.138201e+10]
    -2500.000000: [2348.000000, 1.865899e+10, 2.138201e+10]
    -5000.000000: [2441.000000, 2.174986e+10, 2.311560e+10]
    -7500.000000: [2441.000000, 2.174986e+10, 2.311560e+10]
    -10000.000000: [2546.000000, 2.560064e+10, 2.722188e+10]
    -12500.000000: [2705.000000, 3.228965e+10, 3.433229e+10]
    -15000.000000: [2784.000000, 3.602053e+10, 3.828058e+10]
    -17500.000000: [2784.000000, 3.602053e+10, 3.828058e+10]
    -20000.000000: [2833.000000, 3.844907e+10, 4.085213e+10]
    -22500.000000: [2888.000000, 4.133043e+10, 4.394226e+10]
    -25000.000000: [2939.000000, 4.410815e+10, 4.688483e+10]
    -27500.000000: [3029.000000, 4.929139e+10, 5.235072e+10]
    -30000.000000: [3107.000000, 5.413101e+10, 5.753690e+10]
    -32500.000000: [3119.000000, 5.488824e+10, 5.839891e+10]
    -35000.000000: [3119.000000, 5.488824e+10, 5.839891e+10]
    -37500.000000: [3119.000000, 5.488824e+10, 5.839891e+10]
    -400000.000000: [3144.000000, 5.652157e+10, 6.000752e+10]

Does not work in SeisSol. I attach the error file:
937504.Aqaba.txt
Actually, the file can be parsed with the easi library without a problem (tested with a small CPP program), so it is probably a SeisSol problem.

On the other hand, when I compute explicitly PlasticCo beforehand, then SeisSol does not crash...

!Switch
[rho, mu, lambda]:  !Include Aqaba_rhomulambda.yaml
[bulkFriction]: !ConstantMap
  map:
    bulkFriction: 0.6
[plastCo]: !IdentityMap
  components:
    - !SphericalDomainFilter
      radius: 1800
      centre:
        x: -1540
        y: -22115
        z: -7200
      components: !ConstantMap
        map:
          plastCo: 1.0e20
    - !LayeredModel
          map: !AffineMap
            matrix:
              z1: [0.0, 0.0, 1.0]
            translation:
              z1: 0
          interpolation: lower
          parameters: [plastCo]
          nodes:
            10000.000000: [1.865899e+06]
            -2500.000000: [1.865899e+06]
            -5000.000000: [2.174986e+06]
            -7500.000000: [2.174986e+06]
            -10000.000000: [2.560064e+06]
            -12500.000000: [3.228965e+06]
            -15000.000000: [3.602053e+06]
            -17500.000000: [3.602053e+06]
            -20000.000000: [3.844907e+06]
            -22500.000000: [4.133043e+06]
            -25000.000000: [4.410815e+06]
            -27500.000000: [4.929139e+06]
            -30000.000000: [5.413101e+06]
            -32500.000000: [5.488824e+06]
            -35000.000000: [5.488824e+06]
            -37500.000000: [5.488824e+06]
            -400000.000000: [5.652157e+06]
[s_xx, s_yy, s_zz, s_xy, s_yz, s_xz]: !Include Aqaba_initial_stress.yaml

This is also a problem we experienced with the Ridgecrest setup.

The full setup is available at: /hppfs/work/pr45fi/di73yeq4/Aqaba_bug.

[Thomas-Ulrich]

I could get more info from Sanitizer, on a slightly different setup:

AddressSanitizer:DEADLYSIGNAL
=================================================================
==2074128==ERROR: AddressSanitizer: SEGV on unknown address 0x3c3ef4001fa610 (pc 0x0000004b41ee bp 0x7ffcaacd5a40 sp 0x7ffcaacd59f0 T0)
==2074128==The signal is caused by a WRITE memory access.
    #0 0x4b41ed in easi::ArrayOfStructsAdapter<seissol::model::Plasticity>::set(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, easi::Vector<unsigned int> const&, easi::Slice<double> const&) (/hppfs/work/pr63qo/di73yeq4/SeisSol/build-gcc/SeisSol_Debug_dhsw_4_elastic+0x4b41ed)
    #1 0x46ab2f in easi::Composite::evaluate(easi::Query&, easi::ResultAdapter&) /hppfs/work/pr63qo/di73yeq4/SeisSol/submodules/easi/include/easi/component/Composite.h:113
    #2 0x46acf1 in easi::Composite::evaluate(easi::Query&, easi::ResultAdapter&) /hppfs/work/pr63qo/di73yeq4/SeisSol/submodules/easi/include/easi/component/Composite.h:126
    #3 0x46f87c in easi::EvalModel::evaluate(easi::Query&, easi::ResultAdapter&) /hppfs/work/pr63qo/di73yeq4/SeisSol/submodules/easi/include/easi/component/EvalModel.h:75
    #4 0x46b1a4 in easi::Composite::evaluate(easi::Query&, easi::ResultAdapter&) /hppfs/work/pr63qo/di73yeq4/SeisSol/submodules/easi/include/easi/component/Composite.h:175
    #5 0x46bf13 in easi::Switch::evaluate(easi::Query&, easi::ResultAdapter&) /hppfs/work/pr63qo/di73yeq4/SeisSol/submodules/easi/include/easi/component/Switch.h:105
    #6 0x484298 in seissol::initializers::MaterialParameterDB<seissol::model::Plasticity>::evaluateModel(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, seissol::initializers::QueryGenerator const&) (/hppfs/work/pr63qo/di73yeq4/SeisSol/build-gcc/SeisSol_Debug_dhsw_4_elastic+0x484298)
    #7 0x599439 in seissol::Interoperability::initializeModel(char*, bool, bool, bool, bool, double*, double*, double*, double*, double*) /hppfs/work/pr63qo/di73yeq4/SeisSol/src/Solver/Interoperability.cpp:663
    #8 0x5970a8 in c_interoperability_initializeModel /hppfs/work/pr63qo/di73yeq4/SeisSol/src/Solver/Interoperability.cpp:150
    #9 0x5fe17b in __ini_model_mod_MOD_ini_model /hppfs/work/pr63qo/di73yeq4/SeisSol/src/Physics/ini_model.f90:134
    #10 0x55f484 in __ini_seissol_mod_MOD_ini_seissol /hppfs/work/pr63qo/di73yeq4/SeisSol/src/Initializer/ini_seissol.f90:276
    #11 0x54dba0 in fortran_main /hppfs/work/pr63qo/di73yeq4/SeisSol/src/seissolxx.f90:184
    #12 0x41947a in main /hppfs/work/pr63qo/di73yeq4/SeisSol/src/main.cpp:56
    #13 0x7f0db9b87349 in __libc_start_main (/lib64/libc.so.6+0x24349)
    #14 0x419359 in _start (/hppfs/work/pr63qo/di73yeq4/SeisSol/build-gcc/SeisSol_Debug_dhsw_4_elastic+0x419359)

Here is the material file:

!Switch
[rho, mu, lambda]:  !Include northridge_material.yaml
[bulkFriction]: !ConstantMap
  map:
    bulkFriction: 0.6
#cohesion for off-fault plasticity
[plastCo]: !IdentityMap
  components:
    #turn off plasticity in a sphere around the hypocenter
    - !SphericalDomainFilter
      radius: 2.0e3
      centre:
        x:  24660.0
        y: -6950
        z: -10200.0
      components: !ConstantMap
        map:
          plastCo: 1.0e20
    #depth dependent cohesion
    - !EvalModel
        parameters: [mu]
        model: !Switch
          [mu]: !Include northridge_material.yaml
        components: !FunctionMap
          map:
            plastCo: |
              return 0.0001*mu;
[s_xx, s_yy, s_zz, s_xy, s_yz, s_xz]: !Include Marmara_initial_stress.yaml

Any idea @uphoffc ?

[Thomas-Ulrich]

[plastCo]: !IdentityMap
  components:
    #depth dependent cohesion
    - !EvalModel
        parameters: [mu]
        model: !Switch
          [mu]: !Include northridge_material.yaml
        components: !FunctionMap
          map:
            plastCo: |
              return 0.0001*mu;

This works fine. strange.