Commit aef889ca authored by Jean-Matthieu Gallard's avatar Jean-Matthieu Gallard
Browse files

KernelGen SplitCK vect bug fix

parent a24130bf

KernelGenerator/kernelgenerator/templates/aderdg/fusedSPTVI_linear_split_ck_vect_cpp.template

+14 −16
Original line number Diff line number Diff line
@@ -220,7 +220,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
{% if useFlux %}

    //call flux in x

    for (int yz = 0; yz < {{nDof*nDof3D}}; yz++) {

      solver.{{solverName}}::flux_x_vect(lQi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray); //tmpArray[N][X]

      solver.{{solverName}}::flux_x_vect(lQi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, {{nDofPad}}); //tmpArray[N][X]

      {{ m.matmul('flux_x_sck_vect', 'negativeDudx_by_dx', 'tmpArray', 'lQi_next', '0', '0', idx(0,yz,0,0)) | indent(6) }}{##}

    }

{% endif %}
@@ -243,11 +243,10 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
    //call flux in y

    for (int z = 0; z < {{nDof3D}}; z++) {

      for (int y = 0; y < {{nDof}} ; y++){

          solver.{{solverName}}::flux_y_vect(lQi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+y*{{nDofPad*nVar}}); //tmpArray[Y][N][X]

        solver.{{solverName}}::flux_y_vect(lQi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+y*{{nDofPad*nVar}}, {{nDofPad}}); //tmpArray[Y][N][X]

      }

      //fuse n and x

        {{ m.matmul('flux_y_sck_vect', 'tmpArray', 'negativeDudxT_by_dx', 'lQi_next', '0', '0', idx(z,0,0,0)) | indent(8) }}{##}

      }

      {{ m.matmul('flux_y_sck_vect', 'tmpArray', 'negativeDudxT_by_dx', 'lQi_next', '0', '0', idx(z,0,0,0)) | indent(6) }}{##}

    }

{% endif %}
@@ -271,7 +270,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
    //call flux in z

    for (int y = 0; y < {{nDof}}; y++){

      for (int z = 0; z < {{nDof}}; z++) {

        solver.{{solverName}}::flux_z(lQi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+z*{{nDofPad*nVar}}); //tmpArray[Z][N][X]

        solver.{{solverName}}::flux_z_vect(lQi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+z*{{nDofPad*nVar}}, {{nDofPad}}); //tmpArray[Z][N][X]

      }

      //fuse n and x

      {{ m.matmul('flux_z_sck_vect', 'tmpArray', 'negativeDudxT_by_dx', 'lQi_next', '0', '0', idx(0,y,0,0)) | indent(6) }}{##}
@@ -295,7 +294,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
{% if useSource %}

    //call source

    for (int yz = 0; yz < {{nDof*nDof3D}}; yz++) {

      solver.{{solverName}}::algebraicSource_scalar(lQi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, center, tStep, {{nDofPad}});

      solver.{{solverName}}::algebraicSource_vect(lQi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, center, tStep, {{nDofPad}});

      #pragma omp simd aligned(lQi_next,tmpArray:ALIGNMENT)

      for (int nx = 0; nx < {{nVar*nDofPad}}; nx++) {

        lQi_next[{{idx(0,0,0,nx)}}] -= tmpArray[nx];
@@ -445,7 +444,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
{% if useFlux %}

  // flux in x

  for (int yz = 0; yz < {{nDof*nDof3D}}; yz++) {

    solver.{{solverName}}::flux_x_vect(lQhi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray); //tmpArray[N][X]

    solver.{{solverName}}::flux_x_vect(lQhi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, {{nDofPad}}); //tmpArray[N][X]

    {{ m.matmul('flux_x_sck_vect', 'dudx_by_dx', 'tmpArray', 'lFhi', '0', '0', idx(0,yz,0,0)) | indent(4) }}{##}

  }

{% endif %}
@@ -486,11 +485,10 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
  // flux in y

  for (int z = 0; z < {{nDof3D}}; z++) {

    for (int y = 0; y < {{nDof}} ; y++){

        solver.{{solverName}}::flux_y_vect(lQhi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+y*{{nDofPad*nVar}}); //tmpArray[Y][N][X]

      solver.{{solverName}}::flux_y_vect(lQhi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+y*{{nDofPad*nVar}}, {{nDofPad}}); //tmpArray[Y][N][X]

    }

    //fuse n and x

      {{ m.matmul('flux_y_sck_vect', 'tmpArray', 'dudxT_by_dx', 'lFhi', '0', '0', idx(z,0,0,0)) | indent(6) }}{##}

    }

    {{ m.matmul('flux_y_sck_vect', 'tmpArray', 'dudxT_by_dx', 'lFhi', '0', '0', idx(z,0,0,0)) | indent(4) }}{##}

  }

{% endif %}

  // ncp in y
@@ -536,7 +534,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
  //flux in z

  for (int y = 0; y < {{nDof}}; y++){

    for (int z = 0; z < {{nDof}}; z++) {

      solver.{{solverName}}::flux_z(lQhi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+z*{{nDofPad*nVar}}); //tmpArray[Z][N][X]

      solver.{{solverName}}::flux_z_vect(lQhi+{{idx(z,y,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(z,y,0,0)}}{% else %}nullptr{%endif%}, tmpArray+z*{{nDofPad*nVar}}, {{nDofPad}}); //tmpArray[Z][N][X]

    }

    //fuse n and x

    {{ m.matmul('flux_z_sck_vect', 'tmpArray', 'dudxT_by_dx', 'lFhi', '0', '0', idx(0,y,0,0)) | indent(6) }}{##}
@@ -576,7 +574,7 @@ int {{codeNamespace}}::fusedSpaceTimePredictorVolumeIntegral{{nameSuffix}}( 
{% if useSource %}

    //call source

    for (int yz = 0; yz < {{nDof*nDof3D}}; yz++) {

      solver.{{solverName}}::algebraicSource_scalar(lQhi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, center, tStep, {{nDofPad}});

      solver.{{solverName}}::algebraicSource_vect(lQhi+{{idx(0,yz,0,0)}}, {% if nPar != 0 %}lPi+{{idxLPi(0,yz,0,0)}}{% else %}nullptr{%endif%}, tmpArray, center, tStep, {{nDofPad}});

      for (int x = 0; x < {{nDof}}; x++) {

        for (int n = 0; n < {{nVar}}; n++) {

          lduh[{{idxLduh(0,yz,x,n)}}] -= tmpArray[x*{{nDofPad}}+n] * weights3[{{idxW3(0,yz,x)}}];

LRZ Homepage | Datenschutz | Dokumentation und Betriebsbedingungen | Impressum