Merge branch 'Marc-Precision-handling' into 'development'
Marc precision handling See merge request damask/DAMASK!607
This commit is contained in:
commit
de89d0fe38
|
@ -108,7 +108,7 @@ file(STRINGS "$ENV{PETSC_DIR}/$ENV{PETSC_ARCH}/lib/petsc/conf/petscvariables" PE
|
|||
string(REPLACE "PETSC_FC_INCLUDES = " "" PETSC_INCLUDES "${PETSC_INCLUDES}")
|
||||
message("PETSC_INCLUDES:\n${PETSC_INCLUDES}\n")
|
||||
|
||||
set(CMAKE_Fortran_FLAGS_${CMAKE_BUILD_TYPE} "${BUILDCMD_PRE} ${OPENMP_FLAGS} ${STANDARD_CHECK} ${OPTIMIZATION_FLAGS} ${COMPILE_FLAGS} ${PRECISION_FLAGS}")
|
||||
set(CMAKE_Fortran_FLAGS_${CMAKE_BUILD_TYPE} "${BUILDCMD_PRE} ${OPENMP_FLAGS} ${STANDARD_CHECK} ${OPTIMIZATION_FLAGS} ${COMPILE_FLAGS}")
|
||||
set(CMAKE_Fortran_LINK_EXECUTABLE "${BUILDCMD_PRE} ${CMAKE_Fortran_COMPILER} ${OPENMP_FLAGS} ${OPTIMIZATION_FLAGS} ${LINKER_FLAGS}")
|
||||
|
||||
if(CMAKE_BUILD_TYPE STREQUAL "DEBUG")
|
||||
|
|
|
@ -135,10 +135,3 @@ set (DEBUG_FLAGS "${DEBUG_FLAGS} -fsanitize=undefined")
|
|||
# detect undefined behavior
|
||||
# Additional options
|
||||
# -fsanitize=address,leak,thread
|
||||
|
||||
#------------------------------------------------------------------------------------------------
|
||||
# precision settings
|
||||
set (PRECISION_FLAGS "${PRECISION_FLAGS} -fdefault-real-8")
|
||||
# set precision to 8 bytes for standard real (=8 for pReal). Will set size of double to 16 bytes as long as -fdefault-double-8 is not set
|
||||
set (PRECISION_FLAGS "${PRECISION_FLAGS} -fdefault-double-8")
|
||||
# set precision to 8 bytes for double real, would be 16 bytes if -fdefault-real-8 is used
|
||||
|
|
|
@ -118,8 +118,3 @@ set (DEBUG_FLAGS "${DEBUG_FLAGS} -debug all")
|
|||
# -check: Checks at runtime, where
|
||||
# arg_temp_created: will cause a lot of warnings because we create a bunch of temporary arrays (performance?)
|
||||
# stack:
|
||||
|
||||
#------------------------------------------------------------------------------------------------
|
||||
# precision settings
|
||||
set (PRECISION_FLAGS "${PRECISION_FLAGS} -real-size 64")
|
||||
# set precision for standard real to 32 | 64 | 128 (= 4 | 8 | 16 bytes, type pReal is always 8 bytes)
|
||||
|
|
|
@ -117,8 +117,3 @@ set (DEBUG_FLAGS "${DEBUG_FLAGS} -debug all")
|
|||
# -check: Checks at runtime, where
|
||||
# arg_temp_created: will cause a lot of warnings because we create a bunch of temporary arrays (performance?)
|
||||
# stack:
|
||||
|
||||
#------------------------------------------------------------------------------------------------
|
||||
# precision settings
|
||||
set (PRECISION_FLAGS "${PRECISION_FLAGS} -real-size 64")
|
||||
# set precision for standard real to 32 | 64 | 128 (= 4 | 8 | 16 bytes, type pReal is always 8 bytes)
|
||||
|
|
|
@ -211,7 +211,7 @@ subroutine hypela2(d,g,e,de,s,t,dt,ngens,m,nn,kcus,matus,ndi,nshear,disp, &
|
|||
use OMP_LIB
|
||||
|
||||
implicit none
|
||||
integer, intent(in) :: & ! according to MSC.Marc 2012 Manual D
|
||||
integer(pI64), intent(in) :: & ! according to MSC.Marc 2012 Manual D
|
||||
ngens, & !< size of stress-strain law
|
||||
nn, & !< integration point number
|
||||
ndi, & !< number of direct components
|
||||
|
@ -224,7 +224,7 @@ subroutine hypela2(d,g,e,de,s,t,dt,ngens,m,nn,kcus,matus,ndi,nshear,disp, &
|
|||
jtype, & !< element type
|
||||
ifr, & !< set to 1 if R has been calculated
|
||||
ifu !< set to 1 if stretch has been calculated
|
||||
integer, dimension(2), intent(in) :: & ! according to MSC.Marc 2012 Manual D
|
||||
integer(pI64), dimension(2), intent(in) :: & ! according to MSC.Marc 2012 Manual D
|
||||
m, & !< (1) user element number, (2) internal element number
|
||||
matus, & !< (1) user material identification number, (2) internal material identification number
|
||||
kcus, & !< (1) layer number, (2) internal layer number
|
||||
|
@ -362,7 +362,7 @@ subroutine hypela2(d,g,e,de,s,t,dt,ngens,m,nn,kcus,matus,ndi,nshear,disp, &
|
|||
endif
|
||||
lastLovl = lovl
|
||||
|
||||
call materialpoint_general(computationMode,ffn,ffn1,t(1),timinc,m(1),nn,stress,ddsdde)
|
||||
call materialpoint_general(computationMode,ffn,ffn1,t(1),timinc,int(m(1)),int(nn),stress,ddsdde)
|
||||
|
||||
d = ddsdde(1:ngens,1:ngens)
|
||||
s = stress(1:ndi+nshear)
|
||||
|
@ -383,16 +383,17 @@ subroutine flux(f,ts,n,time)
|
|||
use discretization_Marc
|
||||
|
||||
implicit none
|
||||
real(pReal), dimension(6), intent(in) :: &
|
||||
real(pReal), dimension(6), intent(in) :: &
|
||||
ts
|
||||
integer, dimension(10), intent(in) :: &
|
||||
integer(pI64), dimension(10), intent(in) :: &
|
||||
n
|
||||
real(pReal), intent(in) :: &
|
||||
real(pReal), intent(in) :: &
|
||||
time
|
||||
real(pReal), dimension(2), intent(out) :: &
|
||||
real(pReal), dimension(2), intent(out) :: &
|
||||
f
|
||||
|
||||
f(1) = homogenization_f_T(discretization_Marc_FEM2DAMASK_cell(n(3),n(1)))
|
||||
|
||||
f(1) = homogenization_f_T(discretization_Marc_FEM2DAMASK_cell(int(n(3)),int(n(1))))
|
||||
f(2) = 0.0_pReal
|
||||
|
||||
end subroutine flux
|
||||
|
@ -410,7 +411,8 @@ subroutine uedinc(inc,incsub)
|
|||
use discretization_Marc
|
||||
|
||||
implicit none
|
||||
integer, intent(in) :: inc, incsub
|
||||
integer(pI64), intent(in) :: inc, incsub
|
||||
|
||||
integer :: n, nqncomp, nqdatatype
|
||||
integer, save :: inc_written
|
||||
real(pReal), allocatable, dimension(:,:) :: d_n
|
||||
|
@ -427,9 +429,9 @@ subroutine uedinc(inc,incsub)
|
|||
enddo
|
||||
|
||||
call discretization_Marc_UpdateNodeAndIpCoords(d_n)
|
||||
call materialpoint_results(inc,cptim)
|
||||
call materialpoint_results(int(inc),cptim)
|
||||
|
||||
inc_written = inc
|
||||
inc_written = int(inc)
|
||||
endif
|
||||
|
||||
end subroutine uedinc
|
||||
|
|
|
@ -272,7 +272,7 @@ program DAMASK_grid
|
|||
write(IO_STDOUT,'(2x,a,/,3(3(3x,f12.7,1x)/))',advance='no') 'R:',&
|
||||
transpose(loadCases(l)%rot%asMatrix())
|
||||
|
||||
if (loadCases(l)%r <= 0.0) errorID = 833
|
||||
if (loadCases(l)%r <= 0.0_pReal) errorID = 833
|
||||
if (loadCases(l)%t < 0.0_pReal) errorID = 834
|
||||
if (loadCases(l)%N < 1) errorID = 835
|
||||
if (loadCases(l)%f_out < 1) errorID = 836
|
||||
|
@ -501,7 +501,7 @@ subroutine getMaskedTensor(values,mask,tensor)
|
|||
integer :: i,j
|
||||
|
||||
|
||||
values = 0.0
|
||||
values = 0.0_pReal
|
||||
do i = 1,3
|
||||
row => tensor%get(i)
|
||||
do j = 1,3
|
||||
|
|
|
@ -151,7 +151,7 @@ subroutine VTI_readCellsSizeOrigin(cells,geomSize,origin, &
|
|||
character(len=*), intent(in) :: &
|
||||
fileContent
|
||||
|
||||
character(len=:), allocatable :: dataType, headerType
|
||||
character(len=:), allocatable :: headerType
|
||||
logical :: inFile, inImage, compressed
|
||||
integer(pI64) :: &
|
||||
startPos, endPos
|
||||
|
|
|
@ -229,9 +229,9 @@ pure function cellSurfaceArea(geomSize,cells)
|
|||
real(pReal), dimension(6,1,product(cells)) :: cellSurfaceArea
|
||||
|
||||
|
||||
cellSurfaceArea(1:2,1,:) = geomSize(2)/real(cells(2)) * geomSize(3)/real(cells(3))
|
||||
cellSurfaceArea(3:4,1,:) = geomSize(3)/real(cells(3)) * geomSize(1)/real(cells(1))
|
||||
cellSurfaceArea(5:6,1,:) = geomSize(1)/real(cells(1)) * geomSize(2)/real(cells(2))
|
||||
cellSurfaceArea(1:2,1,:) = geomSize(2)/real(cells(2),pReal) * geomSize(3)/real(cells(3),pReal)
|
||||
cellSurfaceArea(3:4,1,:) = geomSize(3)/real(cells(3),pReal) * geomSize(1)/real(cells(1),pReal)
|
||||
cellSurfaceArea(5:6,1,:) = geomSize(1)/real(cells(1),pReal) * geomSize(2)/real(cells(2),pReal)
|
||||
|
||||
end function cellSurfaceArea
|
||||
|
||||
|
|
|
@ -221,14 +221,14 @@ subroutine grid_mechanical_FEM_init
|
|||
delta = geomSize/real(cells,pReal) ! grid spacing
|
||||
detJ = product(delta) ! cell volume
|
||||
|
||||
BMat = reshape(real([-1.0_pReal/delta(1),-1.0_pReal/delta(2),-1.0_pReal/delta(3), &
|
||||
1.0_pReal/delta(1),-1.0_pReal/delta(2),-1.0_pReal/delta(3), &
|
||||
-1.0_pReal/delta(1), 1.0_pReal/delta(2),-1.0_pReal/delta(3), &
|
||||
1.0_pReal/delta(1), 1.0_pReal/delta(2),-1.0_pReal/delta(3), &
|
||||
-1.0_pReal/delta(1),-1.0_pReal/delta(2), 1.0_pReal/delta(3), &
|
||||
1.0_pReal/delta(1),-1.0_pReal/delta(2), 1.0_pReal/delta(3), &
|
||||
-1.0_pReal/delta(1), 1.0_pReal/delta(2), 1.0_pReal/delta(3), &
|
||||
1.0_pReal/delta(1), 1.0_pReal/delta(2), 1.0_pReal/delta(3)],pReal), [3,8])/4.0_pReal ! shape function derivative matrix
|
||||
BMat = reshape(real([-delta(1)**(-1),-delta(2)**(-1),-delta(3)**(-1), &
|
||||
delta(1)**(-1),-delta(2)**(-1),-delta(3)**(-1), &
|
||||
-delta(1)**(-1), delta(2)**(-1),-delta(3)**(-1), &
|
||||
delta(1)**(-1), delta(2)**(-1),-delta(3)**(-1), &
|
||||
-delta(1)**(-1),-delta(2)**(-1), delta(3)**(-1), &
|
||||
delta(1)**(-1),-delta(2)**(-1), delta(3)**(-1), &
|
||||
-delta(1)**(-1), delta(2)**(-1), delta(3)**(-1), &
|
||||
delta(1)**(-1), delta(2)**(-1), delta(3)**(-1)],pReal), [3,8])/4.0_pReal ! shape function derivative matrix
|
||||
|
||||
HGMat = matmul(transpose(HGcomp),HGcomp) &
|
||||
* HGCoeff*(delta(1)*delta(2) + delta(2)*delta(3) + delta(3)*delta(1))/16.0_pReal ! hourglass stabilization matrix
|
||||
|
@ -656,7 +656,7 @@ subroutine formJacobian(da_local,x_local,Jac_pre,Jac,dummy,err_PETSc)
|
|||
MatNullSpace :: matnull
|
||||
PetscErrorCode :: err_PETSc
|
||||
|
||||
BMatFull = 0.0
|
||||
BMatFull = 0.0_pReal
|
||||
BMatFull(1:3,1 :8 ) = BMat
|
||||
BMatFull(4:6,9 :16) = BMat
|
||||
BMatFull(7:9,17:24) = BMat
|
||||
|
@ -686,7 +686,7 @@ subroutine formJacobian(da_local,x_local,Jac_pre,Jac,dummy,err_PETSc)
|
|||
enddo; enddo; enddo
|
||||
row = col
|
||||
ce = ce + 1
|
||||
K_ele = 0.0
|
||||
K_ele = 0.0_pReal
|
||||
K_ele(1 :8 ,1 :8 ) = HGMat*(homogenization_dPdF(1,1,1,1,ce) + &
|
||||
homogenization_dPdF(2,2,2,2,ce) + &
|
||||
homogenization_dPdF(3,3,3,3,ce))/3.0_pReal
|
||||
|
|
|
@ -198,7 +198,7 @@ subroutine spectral_utilities_init()
|
|||
CHKERRQ(err_PETSc)
|
||||
|
||||
cells1Red = cells(1)/2 + 1
|
||||
wgt = 1.0/real(product(cells),pReal)
|
||||
wgt = real(product(cells),pReal)**(-1)
|
||||
|
||||
num%memory_efficient = num_grid%get_asInt('memory_efficient', defaultVal=1) > 0 ! ToDo: should be logical in YAML file
|
||||
num%divergence_correction = num_grid%get_asInt('divergence_correction', defaultVal=2)
|
||||
|
@ -265,7 +265,7 @@ subroutine spectral_utilities_init()
|
|||
|
||||
N = fftw_mpi_local_size_many(3,[cellsFFTW(3),cellsFFTW(2),cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T],&
|
||||
tensorSize,FFTW_MPI_DEFAULT_BLOCK,PETSC_COMM_WORLD,z,devNull)
|
||||
if (z /= cells3) error stop 'domain decomposition mismatch (tensor)'
|
||||
if (int(z) /= cells3) error stop 'domain decomposition mismatch (tensor)'
|
||||
tensorField = fftw_alloc_complex(N)
|
||||
call c_f_pointer(tensorField,tensorField_real, &
|
||||
[3_C_INTPTR_T,3_C_INTPTR_T,2_C_INTPTR_T*(cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T),cellsFFTW(2),z])
|
||||
|
@ -274,7 +274,7 @@ subroutine spectral_utilities_init()
|
|||
|
||||
N = fftw_mpi_local_size_many(3,[cellsFFTW(3),cellsFFTW(2),cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T],&
|
||||
vectorSize,FFTW_MPI_DEFAULT_BLOCK,PETSC_COMM_WORLD,z,devNull)
|
||||
if (z /= cells3) error stop 'domain decomposition mismatch (vector)'
|
||||
if (int(z) /= cells3) error stop 'domain decomposition mismatch (vector)'
|
||||
vectorField = fftw_alloc_complex(N)
|
||||
call c_f_pointer(vectorField,vectorField_real, &
|
||||
[3_C_INTPTR_T,2_C_INTPTR_T*(cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T),cellsFFTW(2),z])
|
||||
|
@ -283,7 +283,7 @@ subroutine spectral_utilities_init()
|
|||
|
||||
N = fftw_mpi_local_size_3d(cellsFFTW(3),cellsFFTW(2),cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T,&
|
||||
PETSC_COMM_WORLD,z,devNull)
|
||||
if (z /= cells3) error stop 'domain decomposition mismatch (scalar)'
|
||||
if (int(z) /= cells3) error stop 'domain decomposition mismatch (scalar)'
|
||||
scalarField = fftw_alloc_complex(N)
|
||||
call c_f_pointer(scalarField,scalarField_real, &
|
||||
[2_C_INTPTR_T*(cellsFFTW(1)/2_C_INTPTR_T+1_C_INTPTR_T),cellsFFTW(2),z])
|
||||
|
@ -385,17 +385,17 @@ subroutine utilities_updateGamma(C)
|
|||
xiDyad_cmplx(l,m) = conjg(-xi1st(l,i,j,k-cells3Offset))*xi1st(m,i,j,k-cells3Offset)
|
||||
end do
|
||||
do concurrent(l = 1:3, m = 1:3)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal)*xiDyad_cmplx)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal,pReal)*xiDyad_cmplx)
|
||||
end do
|
||||
#else
|
||||
forall(l = 1:3, m = 1:3) &
|
||||
xiDyad_cmplx(l,m) = conjg(-xi1st(l,i,j,k-cells3Offset))*xi1st(m,i,j,k-cells3Offset)
|
||||
forall(l = 1:3, m = 1:3) &
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal)*xiDyad_cmplx)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal,pReal)*xiDyad_cmplx)
|
||||
#endif
|
||||
A(1:3,1:3) = temp33_cmplx%re; A(4:6,4:6) = temp33_cmplx%re
|
||||
A(1:3,4:6) = temp33_cmplx%im; A(4:6,1:3) = -temp33_cmplx%im
|
||||
if (abs(math_det33(A(1:3,1:3))) > 1e-16) then
|
||||
if (abs(math_det33(A(1:3,1:3))) > 1.e-16_pReal) then
|
||||
call math_invert(A_inv, err, A)
|
||||
temp33_cmplx = cmplx(A_inv(1:3,1:3),A_inv(1:3,4:6),pReal)
|
||||
#ifndef __INTEL_COMPILER
|
||||
|
@ -518,17 +518,17 @@ subroutine utilities_fourierGammaConvolution(fieldAim)
|
|||
xiDyad_cmplx(l,m) = conjg(-xi1st(l,i,j,k))*xi1st(m,i,j,k)
|
||||
end do
|
||||
do concurrent(l = 1:3, m = 1:3)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal)*xiDyad_cmplx)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal,pReal)*xiDyad_cmplx)
|
||||
end do
|
||||
#else
|
||||
forall(l = 1:3, m = 1:3) &
|
||||
xiDyad_cmplx(l,m) = conjg(-xi1st(l,i,j,k))*xi1st(m,i,j,k)
|
||||
forall(l = 1:3, m = 1:3) &
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal)*xiDyad_cmplx)
|
||||
temp33_cmplx(l,m) = sum(cmplx(C_ref(l,1:3,m,1:3),0.0_pReal,pReal)*xiDyad_cmplx)
|
||||
#endif
|
||||
A(1:3,1:3) = temp33_cmplx%re; A(4:6,4:6) = temp33_cmplx%re
|
||||
A(1:3,4:6) = temp33_cmplx%im; A(4:6,1:3) = -temp33_cmplx%im
|
||||
if (abs(math_det33(A(1:3,1:3))) > 1e-16) then
|
||||
if (abs(math_det33(A(1:3,1:3))) > 1.e-16_pReal) then
|
||||
call math_invert(A_inv, err, A)
|
||||
temp33_cmplx = cmplx(A_inv(1:3,1:3),A_inv(1:3,4:6),pReal)
|
||||
#ifndef __INTEL_COMPILER
|
||||
|
@ -587,8 +587,8 @@ subroutine utilities_fourierGreenConvolution(D_ref, mu_ref, Delta_t)
|
|||
!$OMP PARALLEL DO PRIVATE(GreenOp_hat)
|
||||
do k = 1, cells3; do j = 1, cells(2) ;do i = 1, cells1Red
|
||||
GreenOp_hat = cmplx(1.0_pReal,0.0_pReal,pReal) &
|
||||
/ (cmplx(mu_ref,0.0_pReal,pReal) + cmplx(Delta_t,0.0_pReal) &
|
||||
* sum(conjg(xi1st(1:3,i,j,k))* matmul(cmplx(D_ref,0.0_pReal),xi1st(1:3,i,j,k))))
|
||||
/ (cmplx(mu_ref,0.0_pReal,pReal) + cmplx(Delta_t,0.0_pReal,pReal) &
|
||||
* sum(conjg(xi1st(1:3,i,j,k))* matmul(cmplx(D_ref,0.0_pReal,pReal),xi1st(1:3,i,j,k))))
|
||||
scalarField_fourier(i,j,k) = scalarField_fourier(i,j,k)*GreenOp_hat
|
||||
enddo; enddo; enddo
|
||||
!$OMP END PARALLEL DO
|
||||
|
@ -608,7 +608,7 @@ real(pReal) function utilities_divergenceRMS()
|
|||
print'(/,1x,a)', '... calculating divergence ................................................'
|
||||
flush(IO_STDOUT)
|
||||
|
||||
rescaledGeom = cmplx(geomSize/scaledGeomSize,0.0_pReal)
|
||||
rescaledGeom = cmplx(geomSize/scaledGeomSize,0.0_pReal,pReal)
|
||||
|
||||
!--------------------------------------------------------------------------------------------------
|
||||
! calculating RMS divergence criterion in Fourier space
|
||||
|
@ -652,7 +652,7 @@ real(pReal) function utilities_curlRMS()
|
|||
print'(/,1x,a)', '... calculating curl ......................................................'
|
||||
flush(IO_STDOUT)
|
||||
|
||||
rescaledGeom = cmplx(geomSize/scaledGeomSize,0.0_pReal)
|
||||
rescaledGeom = cmplx(geomSize/scaledGeomSize,0.0_pReal,pReal)
|
||||
|
||||
!--------------------------------------------------------------------------------------------------
|
||||
! calculating max curl criterion in Fourier space
|
||||
|
|
|
@ -601,9 +601,9 @@ module function RGC_updateState(P,F,avgF,dt,dPdF,ce) result(doneAndHappy)
|
|||
! calculate the stress and penalty due to volume discrepancy
|
||||
vPen = 0.0_pReal
|
||||
do i = 1,nGrain
|
||||
vPen(:,:,i) = -1.0_pReal/real(nGrain,pReal)*num%volDiscrMod*num%volDiscrPow/num%maxVolDiscr* &
|
||||
sign((abs(vDiscrep)/num%maxVolDiscr)**(num%volDiscrPow - 1.0),vDiscrep)* &
|
||||
gVol(i)*transpose(math_inv33(fDef(:,:,i)))
|
||||
vPen(:,:,i) = -real(nGrain,pReal)**(-1)*num%volDiscrMod*num%volDiscrPow/num%maxVolDiscr &
|
||||
* sign((abs(vDiscrep)/num%maxVolDiscr)**(num%volDiscrPow - 1.0_pReal),vDiscrep) &
|
||||
* gVol(i)*transpose(math_inv33(fDef(:,:,i)))
|
||||
end do
|
||||
|
||||
end subroutine volumePenalty
|
||||
|
|
|
@ -2008,7 +2008,7 @@ subroutine buildTransformationSystem(Q,S,Ntrans,cOverA,a_cF,a_cI)
|
|||
],pReal),shape(CFTOHP_SYSTEMTRANS))
|
||||
|
||||
real(pReal), dimension(4,cF_Ntrans), parameter :: &
|
||||
CFTOCI_SYSTEMTRANS = reshape([&
|
||||
CFTOCI_SYSTEMTRANS = real(reshape([&
|
||||
0.0, 1.0, 0.0, 10.26, & ! Pitsch OR (Ma & Hartmaier 2014, Table 3)
|
||||
0.0,-1.0, 0.0, 10.26, &
|
||||
0.0, 0.0, 1.0, 10.26, &
|
||||
|
@ -2021,7 +2021,7 @@ subroutine buildTransformationSystem(Q,S,Ntrans,cOverA,a_cF,a_cI)
|
|||
-1.0, 0.0, 0.0, 10.26, &
|
||||
0.0, 1.0, 0.0, 10.26, &
|
||||
0.0,-1.0, 0.0, 10.26 &
|
||||
],shape(CFTOCI_SYSTEMTRANS))
|
||||
],shape(CFTOCI_SYSTEMTRANS)),pReal)
|
||||
|
||||
integer, dimension(9,cF_Ntrans), parameter :: &
|
||||
CFTOCI_BAINVARIANT = reshape( [&
|
||||
|
@ -2040,7 +2040,7 @@ subroutine buildTransformationSystem(Q,S,Ntrans,cOverA,a_cF,a_cI)
|
|||
],shape(CFTOCI_BAINVARIANT))
|
||||
|
||||
real(pReal), dimension(4,cF_Ntrans), parameter :: &
|
||||
CFTOCI_BAINROT = reshape([&
|
||||
CFTOCI_BAINROT = real(reshape([&
|
||||
1.0, 0.0, 0.0, 45.0, & ! Rotate cF austensite to bain variant
|
||||
1.0, 0.0, 0.0, 45.0, &
|
||||
1.0, 0.0, 0.0, 45.0, &
|
||||
|
@ -2053,7 +2053,7 @@ subroutine buildTransformationSystem(Q,S,Ntrans,cOverA,a_cF,a_cI)
|
|||
0.0, 0.0, 1.0, 45.0, &
|
||||
0.0, 0.0, 1.0, 45.0, &
|
||||
0.0, 0.0, 1.0, 45.0 &
|
||||
],shape(CFTOCI_BAINROT))
|
||||
],shape(CFTOCI_BAINROT)),pReal)
|
||||
|
||||
if (present(a_cI) .and. present(a_cF)) then
|
||||
do i = 1,sum(Ntrans)
|
||||
|
|
|
@ -1110,7 +1110,7 @@ pure function math_rotationalPart(F) result(R)
|
|||
|
||||
C = matmul(transpose(F),F)
|
||||
I_C = math_invariantsSym33(C)
|
||||
I_F = [math_trace33(F), 0.5*(math_trace33(F)**2 - math_trace33(matmul(F,F)))]
|
||||
I_F = [math_trace33(F), 0.5_pReal*(math_trace33(F)**2 - math_trace33(matmul(F,F)))]
|
||||
|
||||
x = math_clip(I_C(1)**2 -3.0_pReal*I_C(2),0.0_pReal)**(3.0_pReal/2.0_pReal)
|
||||
if (dNeq0(x)) then
|
||||
|
@ -1386,7 +1386,7 @@ subroutine selfTest()
|
|||
call random_number(v3_3)
|
||||
call random_number(v3_4)
|
||||
|
||||
if (dNeq(abs(dot_product(math_cross(v3_1-v3_4,v3_2-v3_4),v3_3-v3_4))/6.0, &
|
||||
if (dNeq(abs(dot_product(math_cross(v3_1-v3_4,v3_2-v3_4),v3_3-v3_4))/6.0_pReal, &
|
||||
math_volTetrahedron(v3_1,v3_2,v3_3,v3_4),tol=1.0e-12_pReal)) &
|
||||
error stop 'math_volTetrahedron'
|
||||
|
||||
|
|
|
@ -131,7 +131,7 @@ subroutine FEM_utilities_init
|
|||
call PetscOptionsInsertString(PETSC_NULL_OPTIONS,trim(petsc_optionsOrder),err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
|
||||
wgt = 1.0/real(mesh_maxNips*mesh_NcpElemsGlobal,pReal)
|
||||
wgt = real(mesh_maxNips*mesh_NcpElemsGlobal,pReal)**(-1)
|
||||
|
||||
|
||||
end subroutine FEM_utilities_init
|
||||
|
|
|
@ -247,7 +247,7 @@ subroutine mesh_FEM_build_ipCoordinates(dimPlex,qPoints)
|
|||
mesh_ipCoordinates(dirI,qPt,cell+1) = pV0(dirI)
|
||||
do dirJ = 1, dimPlex
|
||||
mesh_ipCoordinates(dirI,qPt,cell+1) = mesh_ipCoordinates(dirI,qPt,cell+1) + &
|
||||
pCellJ((dirI-1)*dimPlex+dirJ)*(qPoints(qOffset+dirJ) + 1.0)
|
||||
pCellJ((dirI-1)*dimPlex+dirJ)*(qPoints(qOffset+dirJ) + 1.0_pReal)
|
||||
enddo
|
||||
enddo
|
||||
qOffset = qOffset + dimPlex
|
||||
|
|
|
@ -230,14 +230,14 @@ subroutine FEM_mechanical_init(fieldBC)
|
|||
CHKERRQ(err_PETSc)
|
||||
call SNESSetConvergenceTest(mechanical_snes,FEM_mechanical_converged,PETSC_NULL_VEC,PETSC_NULL_FUNCTION,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
call SNESSetTolerances(mechanical_snes,1.0,0.0,0.0,num%itmax,num%itmax,err_PETSc)
|
||||
call SNESSetTolerances(mechanical_snes,1.0_pReal,0.0_pReal,0.0_pReal,num%itmax,num%itmax,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
call SNESSetFromOptions(mechanical_snes,err_PETSc); CHKERRQ(err_PETSc)
|
||||
|
||||
!--------------------------------------------------------------------------------------------------
|
||||
! init fields
|
||||
call VecSet(solution ,0.0,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecSet(solution_rate ,0.0,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecSet(solution ,0.0_pReal,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecSet(solution_rate,0.0_pReal,err_PETSc); CHKERRQ(err_PETSc)
|
||||
allocate(x_scal(cellDof))
|
||||
allocate(nodalWeightsP(1))
|
||||
allocate(nodalPointsP(dimPlex))
|
||||
|
@ -338,11 +338,10 @@ subroutine FEM_mechanical_formResidual(dm_local,xx_local,f_local,dummy,err_PETSc
|
|||
PetscInt :: cellStart, cellEnd, cell, field, face, &
|
||||
qPt, basis, comp, cidx, &
|
||||
numFields, &
|
||||
bcSize,m
|
||||
bcSize,m,i
|
||||
PetscReal :: detFAvg, detJ
|
||||
PetscReal, dimension(dimPlex*dimPlex,cellDof) :: BMat
|
||||
|
||||
IS :: bcPoints
|
||||
IS :: bcPoints
|
||||
|
||||
|
||||
allocate(pV0(dimPlex))
|
||||
|
@ -358,7 +357,7 @@ subroutine FEM_mechanical_formResidual(dm_local,xx_local,f_local,dummy,err_PETSc
|
|||
CHKERRQ(err_PETSc)
|
||||
call DMGetLocalVector(dm_local,x_local,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
call VecWAXPY(x_local,1.0,xx_local,solution_local,err_PETSc)
|
||||
call VecWAXPY(x_local,1.0_pReal,xx_local,solution_local,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
do field = 1, dimPlex; do face = 1, mesh_Nboundaries
|
||||
if (params%fieldBC%componentBC(field)%Mask(face)) then
|
||||
|
@ -386,23 +385,23 @@ subroutine FEM_mechanical_formResidual(dm_local,xx_local,f_local,dummy,err_PETSc
|
|||
IcellJMat = reshape(pInvcellJ,shape=[dimPlex,dimPlex])
|
||||
do qPt = 0, nQuadrature-1
|
||||
m = cell*nQuadrature + qPt+1
|
||||
BMat = 0.0
|
||||
BMat = 0.0_pReal
|
||||
do basis = 0, nBasis-1
|
||||
do comp = 0, dimPlex-1
|
||||
cidx = basis*dimPlex+comp
|
||||
BMat(comp*dimPlex+1:(comp+1)*dimPlex,basis*dimPlex+comp+1) = &
|
||||
matmul(IcellJMat,basisFieldDer((((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp )*dimPlex+1: &
|
||||
(((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp+1)*dimPlex))
|
||||
i = ((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp
|
||||
BMat(comp*dimPlex+1_pPETSCINT:(comp+1_pPETSCINT)*dimPlex,basis*dimPlex+comp+1_pPETSCINT) = &
|
||||
matmul(IcellJMat,basisFieldDer(i*dimPlex+1_pPETSCINT:(i+1_pPETSCINT)*dimPlex))
|
||||
enddo
|
||||
enddo
|
||||
homogenization_F(1:dimPlex,1:dimPlex,m) = reshape(matmul(BMat,x_scal),shape=[dimPlex,dimPlex], order=[2,1])
|
||||
enddo
|
||||
if (num%BBarStabilisation) then
|
||||
detFAvg = math_det33(sum(homogenization_F(1:3,1:3,cell*nQuadrature+1:(cell+1)*nQuadrature),dim=3)/real(nQuadrature))
|
||||
detFAvg = math_det33(sum(homogenization_F(1:3,1:3,cell*nQuadrature+1:(cell+1)*nQuadrature),dim=3)/real(nQuadrature,pReal))
|
||||
do qPt = 0, nQuadrature-1
|
||||
m = cell*nQuadrature + qPt+1
|
||||
homogenization_F(1:dimPlex,1:dimPlex,m) = homogenization_F(1:dimPlex,1:dimPlex,m) &
|
||||
* (detFAvg/math_det33(homogenization_F(1:3,1:3,m)))**(1.0/real(dimPlex))
|
||||
* (detFAvg/math_det33(homogenization_F(1:3,1:3,m)))**(1.0_pReal/real(dimPlex,pReal))
|
||||
|
||||
enddo
|
||||
endif
|
||||
|
@ -425,22 +424,22 @@ subroutine FEM_mechanical_formResidual(dm_local,xx_local,f_local,dummy,err_PETSc
|
|||
call DMPlexComputeCellGeometryAffineFEM(dm_local,cell,pV0,pCellJ,pInvcellJ,detJ,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
IcellJMat = reshape(pInvcellJ,shape=[dimPlex,dimPlex])
|
||||
f_scal = 0.0
|
||||
f_scal = 0.0_pReal
|
||||
do qPt = 0, nQuadrature-1
|
||||
m = cell*nQuadrature + qPt+1
|
||||
BMat = 0.0
|
||||
BMat = 0.0_pReal
|
||||
do basis = 0, nBasis-1
|
||||
do comp = 0, dimPlex-1
|
||||
cidx = basis*dimPlex+comp
|
||||
BMat(comp*dimPlex+1:(comp+1)*dimPlex,basis*dimPlex+comp+1) = &
|
||||
matmul(IcellJMat,basisFieldDer((((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp )*dimPlex+1: &
|
||||
(((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp+1)*dimPlex))
|
||||
i = ((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp
|
||||
BMat(comp*dimPlex+1_pPETSCINT:(comp+1_pPETSCINT)*dimPlex,basis*dimPlex+comp+1_pPETSCINT) = &
|
||||
matmul(IcellJMat,basisFieldDer(i*dimPlex+1_pPETSCINT:(i+1_pPETSCINT)*dimPlex))
|
||||
enddo
|
||||
enddo
|
||||
f_scal = f_scal + &
|
||||
matmul(transpose(BMat), &
|
||||
reshape(transpose(homogenization_P(1:dimPlex,1:dimPlex,m)), &
|
||||
shape=[dimPlex*dimPlex]))*qWeights(qPt+1)
|
||||
f_scal = f_scal &
|
||||
+ matmul(transpose(BMat), &
|
||||
reshape(transpose(homogenization_P(1:dimPlex,1:dimPlex,m)), &
|
||||
shape=[dimPlex*dimPlex]))*qWeights(qPt+1_pPETSCINT)
|
||||
enddo
|
||||
f_scal = f_scal*abs(detJ)
|
||||
pf_scal => f_scal
|
||||
|
@ -465,28 +464,25 @@ subroutine FEM_mechanical_formJacobian(dm_local,xx_local,Jac_pre,Jac,dummy,err_P
|
|||
PetscObject, intent(in) :: dummy
|
||||
PetscErrorCode :: err_PETSc
|
||||
|
||||
PetscDS :: prob
|
||||
Vec :: x_local, xx_local
|
||||
|
||||
PetscSection :: section, gSection
|
||||
PetscDS :: prob
|
||||
Vec :: x_local, xx_local
|
||||
PetscSection :: section, gSection
|
||||
|
||||
PetscReal, dimension(1, cellDof) :: MatB
|
||||
PetscReal, dimension(dimPlex**2,cellDof) :: BMat, BMatAvg, MatA
|
||||
PetscReal, dimension(3,3) :: F, FAvg, FInv
|
||||
PetscReal :: detJ
|
||||
PetscReal, dimension(:), pointer :: basisField, basisFieldDer, &
|
||||
PetscReal, dimension(3,3) :: F, FAvg, FInv
|
||||
PetscReal :: detJ
|
||||
PetscReal, dimension(:), pointer :: basisField, basisFieldDer, &
|
||||
pV0, pCellJ, pInvcellJ
|
||||
|
||||
PetscScalar, dimension(:), pointer :: pK_e, x_scal
|
||||
|
||||
PetscScalar,dimension(cellDOF,cellDOF), target :: K_e
|
||||
PetscScalar,dimension(cellDOF,cellDOF) :: K_eA , &
|
||||
K_eB
|
||||
PetscScalar,dimension(cellDOF,cellDOF) :: K_eA, K_eB
|
||||
|
||||
PetscInt :: cellStart, cellEnd, cell, field, face, &
|
||||
qPt, basis, comp, cidx,bcSize, m
|
||||
|
||||
IS :: bcPoints
|
||||
PetscInt :: cellStart, cellEnd, cell, field, face, &
|
||||
qPt, basis, comp, cidx,bcSize, m, i
|
||||
IS :: bcPoints
|
||||
|
||||
|
||||
allocate(pV0(dimPlex))
|
||||
|
@ -530,30 +526,29 @@ subroutine FEM_mechanical_formJacobian(dm_local,xx_local,Jac_pre,Jac,dummy,err_P
|
|||
CHKERRQ(err_PETSc)
|
||||
call DMPlexComputeCellGeometryAffineFEM(dm_local,cell,pV0,pCellJ,pInvcellJ,detJ,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
K_eA = 0.0
|
||||
K_eB = 0.0
|
||||
MatB = 0.0
|
||||
FAvg = 0.0
|
||||
BMatAvg = 0.0
|
||||
K_eA = 0.0_pReal
|
||||
K_eB = 0.0_pReal
|
||||
MatB = 0.0_pReal
|
||||
FAvg = 0.0_pReal
|
||||
BMatAvg = 0.0_pReal
|
||||
do qPt = 0, nQuadrature-1
|
||||
m = cell*nQuadrature + qPt + 1
|
||||
BMat = 0.0
|
||||
BMat = 0.0_pReal
|
||||
do basis = 0, nBasis-1
|
||||
do comp = 0, dimPlex-1
|
||||
cidx = basis*dimPlex+comp
|
||||
BMat(comp*dimPlex+1:(comp+1)*dimPlex,basis*dimPlex+comp+1) = &
|
||||
matmul( reshape(pInvcellJ, shape = [dimPlex,dimPlex]),&
|
||||
basisFieldDer((((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp )*dimPlex+1: &
|
||||
(((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp+1)*dimPlex))
|
||||
i = ((qPt*nBasis + basis)*dimPlex + comp)*dimPlex+comp
|
||||
BMat(comp*dimPlex+1_pPETSCINT:(comp+1_pPETSCINT)*dimPlex,basis*dimPlex+comp+1_pPETSCINT) = &
|
||||
matmul(reshape(pInvcellJ,[dimPlex,dimPlex]),basisFieldDer(i*dimPlex+1_pPETSCINT:(i+1_pPETSCINT)*dimPlex))
|
||||
enddo
|
||||
enddo
|
||||
MatA = matmul(reshape(reshape(homogenization_dPdF(1:dimPlex,1:dimPlex,1:dimPlex,1:dimPlex,m), &
|
||||
shape=[dimPlex,dimPlex,dimPlex,dimPlex], order=[2,1,4,3]), &
|
||||
shape=[dimPlex*dimPlex,dimPlex*dimPlex]),BMat)*qWeights(qPt+1)
|
||||
shape=[dimPlex*dimPlex,dimPlex*dimPlex]),BMat)*qWeights(qPt+1_pPETSCINT)
|
||||
if (num%BBarStabilisation) then
|
||||
F(1:dimPlex,1:dimPlex) = reshape(matmul(BMat,x_scal),shape=[dimPlex,dimPlex])
|
||||
FInv = math_inv33(F)
|
||||
K_eA = K_eA + matmul(transpose(BMat),MatA)*math_det33(FInv)**(1.0/real(dimPlex))
|
||||
K_eA = K_eA + matmul(transpose(BMat),MatA)*math_det33(FInv)**(1.0_pReal/real(dimPlex,pReal))
|
||||
K_eB = K_eB - &
|
||||
matmul(transpose(matmul(reshape(homogenization_F(1:dimPlex,1:dimPlex,m),shape=[dimPlex**2,1_pPETSCINT]), &
|
||||
matmul(reshape(FInv(1:dimPlex,1:dimPlex), &
|
||||
|
@ -568,10 +563,10 @@ subroutine FEM_mechanical_formJacobian(dm_local,xx_local,Jac_pre,Jac,dummy,err_P
|
|||
enddo
|
||||
if (num%BBarStabilisation) then
|
||||
FInv = math_inv33(FAvg)
|
||||
K_e = K_eA*math_det33(FAvg/real(nQuadrature))**(1.0/real(dimPlex)) + &
|
||||
K_e = K_eA*math_det33(FAvg/real(nQuadrature,pReal))**(1.0_pReal/real(dimPlex,pReal)) + &
|
||||
(matmul(matmul(transpose(BMatAvg), &
|
||||
reshape(FInv(1:dimPlex,1:dimPlex),shape=[dimPlex**2,1_pPETSCINT],order=[2,1])),MatB) + &
|
||||
K_eB)/real(dimPlex)
|
||||
K_eB)/real(dimPlex,pReal)
|
||||
else
|
||||
K_e = K_eA
|
||||
endif
|
||||
|
@ -641,7 +636,7 @@ subroutine FEM_mechanical_forward(guess,timeinc,timeinc_old,fieldBC)
|
|||
CHKERRQ(err_PETSc)
|
||||
call DMGlobalToLocalEnd(dm_local,solution,INSERT_VALUES,x_local,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
call VecAXPY(solution_local,1.0,x_local,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecAXPY(solution_local,1.0_pReal,x_local,err_PETSc); CHKERRQ(err_PETSc)
|
||||
do field = 1, dimPlex; do face = 1, mesh_Nboundaries
|
||||
if (fieldBC%componentBC(field)%Mask(face)) then
|
||||
call DMGetStratumSize(dm_local,'Face Sets',mesh_boundaries(face),bcSize,err_PETSc)
|
||||
|
@ -659,7 +654,7 @@ subroutine FEM_mechanical_forward(guess,timeinc,timeinc_old,fieldBC)
|
|||
!--------------------------------------------------------------------------------------------------
|
||||
! update rate and forward last inc
|
||||
call VecCopy(solution,solution_rate,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecScale(solution_rate,1.0/timeinc_old,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecScale(solution_rate,timeinc_old**(-1),err_PETSc); CHKERRQ(err_PETSc)
|
||||
endif
|
||||
call VecCopy(solution_rate,solution,err_PETSc); CHKERRQ(err_PETSc)
|
||||
call VecScale(solution,timeinc,err_PETSc); CHKERRQ(err_PETSc)
|
||||
|
@ -685,9 +680,8 @@ subroutine FEM_mechanical_converged(snes_local,PETScIter,xnorm,snorm,fnorm,reaso
|
|||
call SNESConvergedDefault(snes_local,PETScIter,xnorm,snorm,fnorm/divTol,reason,dummy,err_PETSc)
|
||||
CHKERRQ(err_PETSc)
|
||||
if (terminallyIll) reason = SNES_DIVERGED_FUNCTION_DOMAIN
|
||||
print'(/,1x,a,a,i0,a,i0,f0.3)', trim(incInfo), &
|
||||
' @ Iteration ',PETScIter,' mechanical residual norm = ', &
|
||||
int(fnorm/divTol),fnorm/divTol-int(fnorm/divTol) ! ToDo: int casting?
|
||||
print'(/,1x,a,a,i0,a,f0.3)', trim(incInfo), &
|
||||
' @ Iteration ',PETScIter,' mechanical residual norm = ',fnorm/divTol
|
||||
print'(/,1x,a,/,2(3(2x,f12.4,1x)/),3(2x,f12.4,1x))', &
|
||||
'Piola--Kirchhoff stress / MPa =',transpose(P_av)*1.e-6_pReal
|
||||
flush(IO_STDOUT)
|
||||
|
|
|
@ -593,7 +593,7 @@ function integrateStateFPI(F_0,F,subFp0,subFi0,subState0,Delta_t,ph,en) result(b
|
|||
|
||||
iteration: do NiterationState = 1, num%nState
|
||||
|
||||
dotState_last(1:sizeDotState,2) = merge(dotState_last(1:sizeDotState,1),0.0, nIterationState > 1)
|
||||
dotState_last(1:sizeDotState,2) = merge(dotState_last(1:sizeDotState,1),0.0_pReal, nIterationState > 1)
|
||||
dotState_last(1:sizeDotState,1) = dotState
|
||||
|
||||
broken = integrateStress(F,subFp0,subFi0,Delta_t,ph,en)
|
||||
|
@ -756,7 +756,7 @@ function integrateStateRK4(F_0,F,subFp0,subFi0,subState0,Delta_t,ph,en) result(b
|
|||
real(pReal), dimension(3), parameter :: &
|
||||
C = [0.5_pReal, 0.5_pReal, 1.0_pReal]
|
||||
real(pReal), dimension(4), parameter :: &
|
||||
B = [1.0_pReal/6.0_pReal, 1.0_pReal/3.0_pReal, 1.0_pReal/3.0_pReal, 1.0_pReal/6.0_pReal]
|
||||
B = [6.0_pReal, 3.0_pReal, 3.0_pReal, 6.0_pReal]**(-1)
|
||||
|
||||
|
||||
broken = integrateStateRK(F_0,F,subFp0,subFi0,subState0,Delta_t,ph,en,A,B,C)
|
||||
|
@ -1263,8 +1263,6 @@ module subroutine mechanical_restartRead(groupHandle,ph)
|
|||
integer(HID_T), intent(in) :: groupHandle
|
||||
integer, intent(in) :: ph
|
||||
|
||||
integer :: en
|
||||
|
||||
|
||||
call HDF5_read(plasticState(ph)%state0,groupHandle,'omega_plastic')
|
||||
call HDF5_read(phase_mechanical_S0(ph)%data,groupHandle,'S')
|
||||
|
|
|
@ -912,7 +912,7 @@ pure subroutine kinetics_tw(Mp,T,abs_dot_gamma_sl,ph,en,&
|
|||
real(pReal), dimension(param(ph)%sum_N_tw), optional, intent(out) :: &
|
||||
ddot_gamma_dtau_tw
|
||||
|
||||
real :: &
|
||||
real(pReal) :: &
|
||||
tau, tau_r, tau_hat, &
|
||||
dot_N_0, &
|
||||
x0, V, &
|
||||
|
@ -988,7 +988,7 @@ pure subroutine kinetics_tr(Mp,T,abs_dot_gamma_sl,ph,en,&
|
|||
real(pReal), dimension(param(ph)%sum_N_tr), optional, intent(out) :: &
|
||||
ddot_gamma_dtau_tr
|
||||
|
||||
real :: &
|
||||
real(pReal) :: &
|
||||
tau, tau_r, tau_hat, &
|
||||
dot_N_0, &
|
||||
x0, V, &
|
||||
|
|
|
@ -801,7 +801,7 @@ subroutine selfTest()
|
|||
real(pReal), dimension(3,3) :: om, t33
|
||||
real(pReal), dimension(3,3,3,3) :: t3333
|
||||
real(pReal), dimension(6,6) :: C
|
||||
real :: A,B
|
||||
real(pReal) :: A,B
|
||||
integer :: i
|
||||
|
||||
|
||||
|
|
Loading…
Reference in New Issue