223031012
/
DAMASK_EICMD
mirror of https://github.com/achalhp/DAMASK_EICMD.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
DAMASK_EICMD/code/DAMASK_spectral_utilities.f90

982 lines
56 KiB
Fortran
Raw Normal View History

updated copyright header
2013-03-22 23:05:05 +05:30
! Copyright 2011-13 Max-Planck-Institut für Eisenforschung GmbH
!
! This file is part of DAMASK,
! the Düsseldorf Advanced MAterial Simulation Kit.
!
! DAMASK is free software: you can redistribute it and/or modify
! it under the terms of the GNU General Public License as published by
! the Free Software Foundation, either version 3 of the License, or
! (at your option) any later version.
!
! DAMASK is distributed in the hope that it will be useful,
! but WITHOUT ANY WARRANTY; without even the implied warranty of
! MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
! GNU General Public License for more details.
!
! You should have received a copy of the GNU General Public License
! along with DAMASK. If not, see <http://www.gnu.org/licenses/>.
!
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
! $Id$
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
!> @author Pratheek Shanthraj, Max-Planck-Institut für Eisenforschung GmbH
!> @author Martin Diehl, Max-Planck-Institut für Eisenforschung GmbH
!> @author Philip Eisenlohr, Max-Planck-Institut für Eisenforschung GmbH
!> @brief Utilities used by the different spectral solver variants
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
module DAMASK_spectral_utilities
use, intrinsic :: iso_c_binding
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
use prec, only: &
pReal, &
pInt
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
implicit none
added missing line continuation in numerics and explicit private statements in the other files
2013-03-28 16:07:00 +05:30
private
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
#ifdef PETSc
#include <finclude/petscsys.h>
#endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
logical, public :: cutBack =.false. !< cut back of BVP solver in case convergence is not achieved or a material point is terminally ill
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
!--------------------------------------------------------------------------------------------------
! grid related information information
integer(pInt), public, dimension(3) :: grid !< grid points as specified in geometry file
real(pReal), public :: wgt !< weighting factor 1/Nelems
real(pReal), public, dimension(3) :: geomSize !< size of geometry as specified in geometry file
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
!--------------------------------------------------------------------------------------------------
! variables storing information for spectral method and FFTW
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
integer(pInt), public :: grid1Red !< grid(1)/2
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
real(pReal), public, dimension(:,:,:,:,:), pointer :: field_real !< real representation (some stress or deformation) of field_fourier
complex(pReal),private, dimension(:,:,:,:,:), pointer :: field_fourier !< field on which the Fourier transform operates
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), private, dimension(:,:,:,:,:,:,:), allocatable :: gamma_hat !< gamma operator (field) for spectral method
real(pReal), private, dimension(:,:,:,:), allocatable :: xi !< wave vector field for divergence and for gamma operator
real(pReal), private, dimension(3,3,3,3) :: C_ref !< reference stiffness
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), private, dimension(3) :: scaledGeomSize !< scaled geometry size for calculation of divergence (Basic, Basic PETSc)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
!--------------------------------------------------------------------------------------------------
! debug fftw
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
complex(pReal),private, dimension(:,:,:), pointer :: scalarField_real, & !< scalar field real representation for debug of FFTW
scalarField_fourier !< scalar field complex representation for debug of FFTW
intermediate, not working state of new solver structure
2012-07-20 21:03:13 +05:30
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
!--------------------------------------------------------------------------------------------------
! debug divergence
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
real(pReal), private, dimension(:,:,:,:), pointer :: divReal !< scalar field real representation for debugging divergence calculation
complex(pReal),private, dimension(:,:,:,:), pointer :: divFourier !< scalar field real representation for debugging divergence calculation
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
!--------------------------------------------------------------------------------------------------
! plans for FFTW
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
type(C_PTR), private :: &
planForth, & !< FFTW plan P(x) to P(k)
planBack, & !< FFTW plan F(k) to F(x)
planDebugForth, & !< FFTW plan for scalar field (proof that order of usual transform is correct)
planDebugBack, & !< FFTW plan for scalar field inverse (proof that order of usual transform is correct)
planDiv !< plan for FFTW in case of debugging divergence calculation
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
! variables controlling debugging
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
logical, private :: &
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
debugGeneral, & !< general debugging of spectral solver
debugDivergence, & !< debugging of divergence calculation (comparison to function used for post processing)
simplified the rate calculation interface in DAMASK_spectral_utilities.f90, corrected bug in rotation parsing and added debug option "rotation" to spectral solver options. This will show the current average stress and deformation additionally in lab coordinate system
2012-11-28 20:34:05 +05:30
debugFFTW, & !< doing additional FFT on scalar field and compare to results of strided 3D FFT
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
debugRotation, & !< also printing out results in lab frame
debugPETSc !< use some in debug defined options for more verbose PETSc solution
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
next round in modularization
2012-07-25 19:31:39 +05:30
!--------------------------------------------------------------------------------------------------
new version of modular solver structure
2012-08-03 14:55:48 +05:30
! derived types
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
type, public :: tSolutionState !< return type of solution from spectral solver variants
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
logical :: converged = .true.
logical :: regrid = .false.
logical :: termIll = .false.
integer(pInt) :: iterationsNeeded = 0_pInt
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
end type tSolutionState
new version of modular solver structure
2012-08-03 14:55:48 +05:30
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
type, public :: tBoundaryCondition !< set of parameters defining a boundary condition
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), dimension(3,3) :: values = 0.0_pReal
real(pReal), dimension(3,3) :: maskFloat = 0.0_pReal
new version of modular solver structure
2012-08-03 14:55:48 +05:30
logical, dimension(3,3) :: maskLogical = .false.
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
character(len=64) :: myType = 'None'
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
end type tBoundaryCondition
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
public :: &
utilities_init, &
utilities_updateGamma, &
utilities_FFTforward, &
utilities_FFTbackward, &
utilities_fourierConvolution, &
utilities_divergenceRMS, &
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
utilities_curlRMS, &
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_maskedCompliance, &
utilities_constitutiveResponse, &
utilities_calculateRate, &
utilities_forwardField, &
utilities_destroy
private :: &
utilities_getFilter
new version of modular solver structure
2012-08-03 14:55:48 +05:30
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
contains
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
!> @brief allocates all neccessary fields, sets debug flags, create plans for FFTW
!> @details Sets the debug levels for general, divergence, restart and FFTW from the biwise coding
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
!> provided by the debug module to logicals.
!> Allocates all fields used by FFTW and create the corresponding plans depending on the debug
!> level chosen.
!> Initializes FFTW.
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
subroutine utilities_init()
moved some more 'mesh related' functions for post processing from math.f90 to mesh.f90 f2py functions remaining in math.f90 now uses assumed size arrays in order to have simpler interfaces. This is only working with python 2.7! changed python pre- and postprocessing scripts. If you encounter any problems whith core modules, try to remove the old core.so in the lib/damask
2012-08-27 13:34:47 +05:30
use, intrinsic :: iso_fortran_env ! to get compiler_version and compiler_options (at least for gfortran >4.6 at the moment)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
use DAMASK_interface, only: &
geometryFile
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use IO, only: &
added some warning and explicit size of arrays
2013-01-10 19:03:43 +05:30
IO_error, &
reporting current time for most module inits
2013-02-25 22:04:59 +05:30
IO_warning, &
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
IO_timeStamp, &
IO_open_file
new version of modular solver structure
2012-08-03 14:55:48 +05:30
use numerics, only: &
intermediate, not working state of new solver structure
2012-07-20 21:03:13 +05:30
DAMASK_NumThreadsInt, &
fftw_planner_flag, &
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
fftw_timelimit, &
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
memory_efficient, &
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
petsc_options, &
divergence_correction
intermediate, not working state of new solver structure
2012-07-20 21:03:13 +05:30
use debug, only: &
debug_level, &
corrected PETSc debug and Capitalization
2013-05-16 14:25:10 +05:30
debug_SPECTRAL, &
debug_LEVELBASIC, &
debug_SPECTRALDIVERGENCE, &
debug_SPECTRALFFTW, &
debug_SPECTRALPETSC, &
debug_SPECTRALROTATION
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
#ifdef PETSc
added funtionality to specify working directory to spectral solver
2013-01-02 22:32:12 +05:30
use debug, only: &
corrected PETSc debug and Capitalization
2013-05-16 14:25:10 +05:30
PETSCDEBUG
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
#endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math ! must use the whole module for use of FFTW
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
use mesh, only: &
mesh_spectral_getSize, &
mesh_spectral_getGrid
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
implicit none
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
#ifdef PETSc
updated doxygen config file for FEM solvers, replaced some external statements, removed debugging options from abaqus_v6.env causing Abaqus to crash (didn't detect the bug so far), added new homogenization test to automator, added more debug options + comments to Makefile
2013-02-13 23:24:56 +05:30
external :: &
PETScOptionsClear, &
PETScOptionsInsertString, &
MPI_Abort
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
PetscErrorCode :: ierr
#endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
integer(pInt) :: i, j, k
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
integer(pInt), parameter :: fileUnit = 228_pInt
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
integer(pInt), dimension(3) :: k_s
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
type(C_PTR) :: &
tensorField, & !< field cotaining data for FFTW in real and fourier space (in place)
scalarField_realC, & !< field cotaining data for FFTW in real space when debugging FFTW (no in place)
scalarField_fourierC, & !< field cotaining data for FFTW in fourier space when debugging FFTW (no in place)
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
div !< field cotaining data for FFTW in real and fourier space when debugging divergence (in place)
fixed bug in coordinate calculation when no average F was given (scaled wrong by ncp_elems**2) other changes: just polishing + some more comments
2013-03-27 17:58:55 +05:30
write(6,'(/,a)') ' <<<+- DAMASK_spectral_utilities init -+>>>'
write(6,'(a)') ' $Id$'
write(6,'(a15,a)') ' Current time: ',IO_timeStamp()
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
#include "compilation_info.f90"
!--------------------------------------------------------------------------------------------------
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
! set debugging parameters
corrected PETSc debug and Capitalization
2013-05-16 14:25:10 +05:30
debugGeneral = iand(debug_level(debug_SPECTRAL),debug_LEVELBASIC) /= 0
debugDivergence = iand(debug_level(debug_SPECTRAL),debug_SPECTRALDIVERGENCE) /= 0
debugFFTW = iand(debug_level(debug_SPECTRAL),debug_SPECTRALFFTW) /= 0
debugRotation = iand(debug_level(debug_SPECTRAL),debug_SPECTRALROTATION) /= 0
debugPETSc = iand(debug_level(debug_SPECTRAL),debug_SPECTRALPETSC) /= 0
added some warning and explicit size of arrays
2013-01-10 19:03:43 +05:30
#ifdef PETSc
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
if(debugPETSc) write(6,'(/,a)') ' Initializing PETSc with debug options: ', trim(PETScDebug), &
' add more using the PETSc_Options keyword in numerics.config '
flush(6)
call PetscOptionsClear(ierr); CHKERRQ(ierr)
corrected PETSc debug and Capitalization
2013-05-16 14:25:10 +05:30
if(debugPETSc) call PetscOptionsInsertString(trim(PETSCDEBUG),ierr); CHKERRQ(ierr)
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
call PetscOptionsInsertString(trim(petsc_options),ierr); CHKERRQ(ierr)
added some warning and explicit size of arrays
2013-01-10 19:03:43 +05:30
#else
corrected PETSc debug and Capitalization
2013-05-16 14:25:10 +05:30
if(debugPETSc) call IO_warning(41_pInt, ext_msg='debug PETSc')
made PETSc silent by removing -snes_view from default options. introduced PETSc option for debugging that introduces some debugging options into the petsc options and move PETSc initialization from numerics to DAMASK_spectral_utilities
2012-12-15 23:37:49 +05:30
#endif
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
call IO_open_file(fileUnit,geometryFile) ! parse info from geometry file...
grid = mesh_spectral_getGrid(fileUnit)
grid1Red = grid(1)/2_pInt + 1_pInt
wgt = 1.0/real(product(grid),pReal)
geomSize = mesh_spectral_getSize(fileUnit)
close(fileUnit)
write(6,'(a,3(i12 ))') ' grid a b c: ', grid
write(6,'(a,3(f12.5))') ' size x y z: ', geomSize
!--------------------------------------------------------------------------------------------------
! scale dimension to calculate either uncorrected, dimension-independent, or dimension- and reso-
! lution-independent divergence
if (divergence_correction == 1_pInt) then
do j = 1_pInt, 3_pInt
if (j /= minloc(geomSize,1) .and. j /= maxloc(geomSize,1)) &
scaledGeomSize = geomSize/geomSize(j)
enddo
elseif (divergence_correction == 2_pInt) then
do j = 1_pInt, 3_pInt
if (j /= minloc(geomSize/grid,1) .and. j /= maxloc(geomSize/grid,1)) &
scaledGeomSize = geomSize/geomSize(j)*grid(j)
enddo
else
scaledGeomSize = geomSize
endif
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
! allocation
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
allocate (xi(3,grid1Red,grid(2),grid(3)),source = 0.0_pReal) ! frequencies, only half the size for first dimension
tensorField = fftw_alloc_complex(int(grid1Red*grid(2)*grid(3)*9_pInt,C_SIZE_T)) ! allocate aligned data using a C function, C_SIZE_T is of type integer(8)
call c_f_pointer(tensorField, field_real, [grid(1)+2_pInt-mod(grid(1),2_pInt),grid(2),grid(3),3,3])! place a pointer for a real representation on tensorField
call c_f_pointer(tensorField, field_fourier,[grid1Red, grid(2),grid(3),3,3])! place a pointer for a complex representation on tensorField
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
! general initialization of FFTW (see manual on fftw.org for more details)
made restart working for all solvers and added corresponding test corrected error handling for FFT based function in case of odd numbers etc.
2012-12-14 23:00:22 +05:30
if (pReal /= C_DOUBLE .or. pInt /= C_INT) call IO_error(0_pInt,ext_msg='Fortran to C') ! check for correct precision in C
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!$ if(DAMASK_NumThreadsInt > 0_pInt) then
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!$ i = fftw_init_threads() ! returns 0 in case of problem
!$ if (i == 0_pInt) call IO_error(error_ID = 809_pInt)
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!$ call fftw_plan_with_nthreads(DAMASK_NumThreadsInt)
!$ endif
call fftw_set_timelimit(fftw_timelimit) ! set timelimit for plan creation
!--------------------------------------------------------------------------------------------------
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
! creating plans for the convolution
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
planForth = fftw_plan_many_dft_r2c(3,[grid(3),grid(2) ,grid(1)], 9, & ! dimensions, logical length in each dimension in reversed order, no. of transforms
field_real,[grid(3),grid(2) ,grid(1)+2_pInt-mod(grid(1),2_pInt)], & ! input data, physical length in each dimension in reversed order
1, grid(3)*grid(2)*(grid(1)+2_pInt-mod(grid(1),2_pInt)), & ! striding, product of physical length in the 3 dimensions
field_fourier,[grid(3),grid(2) ,grid1Red], & ! output data, physical length in each dimension in reversed order
1, grid(3)*grid(2)* grid1Red, fftw_planner_flag) ! striding, product of physical length in the 3 dimensions, planner precision
planBack = fftw_plan_many_dft_c2r(3,[grid(3),grid(2) ,grid(1)], 9, & ! dimensions, logical length in each dimension in reversed order, no. of transforms
field_fourier,[grid(3),grid(2) ,grid1Red], & ! input data, physical length in each dimension in reversed order
1, grid(3)*grid(2)* grid1Red, & ! striding, product of physical length in the 3 dimensions
field_real,[grid(3),grid(2) ,grid(1)+2_pInt-mod(grid(1),2_pInt)], & ! output data, physical length in each dimension in reversed order
1, grid(3)*grid(2)*(grid(1)+2_pInt-mod(grid(1),2_pInt)), & ! striding, product of physical length in the 3 dimensions
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
fftw_planner_flag) ! planner precision
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
! depending on debug options, allocate more memory and create additional plans
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
if (debugDivergence) then
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
div = fftw_alloc_complex(int(grid1Red*grid(2)*grid(3)*3_pInt,C_SIZE_T))
call c_f_pointer(div,divReal, [grid(1)+2_pInt-mod(grid(1),2_pInt),grid(2),grid(3),3])
call c_f_pointer(div,divFourier,[grid1Red, grid(2),grid(3),3])
planDiv = fftw_plan_many_dft_c2r(3,[grid(3),grid(2) ,grid(1)],3,&
divFourier,[grid(3),grid(2) ,grid1Red],&
1, grid(3)*grid(2)* grid1Red,&
divReal,[grid(3),grid(2) ,grid(1)+2_pInt-mod(grid(1),2_pInt)], &
1, grid(3)*grid(2)*(grid(1)+2_pInt-mod(grid(1),2_pInt)), &
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
fftw_planner_flag)
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
endif
if (debugFFTW) then
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
scalarField_realC = fftw_alloc_complex(int(product(grid),C_SIZE_T)) ! allocate data for real representation (no in place transform)
scalarField_fourierC = fftw_alloc_complex(int(product(grid),C_SIZE_T)) ! allocate data for fourier representation (no in place transform)
call c_f_pointer(scalarField_realC, scalarField_real, grid) ! place a pointer for a real representation
call c_f_pointer(scalarField_fourierC, scalarField_fourier, grid) ! place a pointer for a fourier representation
planDebugForth = fftw_plan_dft_3d(grid(3),grid(2),grid(1),& ! reversed order (C style)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
scalarField_real,scalarField_fourier,-1,fftw_planner_flag) ! input, output, forward FFT(-1), planner precision
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
planDebugBack = fftw_plan_dft_3d(grid(3),grid(2),grid(1),& ! reversed order (C style)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
scalarField_fourier,scalarField_real,+1,fftw_planner_flag) ! input, output, backward (1), planner precision
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
endif
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
if (debugGeneral) write(6,'(/,a)') ' FFTW initialized'
flush(6)
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
! calculation of discrete angular frequencies, ordered as in FFTW (wrap around)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3)
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
k_s(3) = k - 1_pInt
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(k > grid(3)/2_pInt + 1_pInt) k_s(3) = k_s(3) - grid(3) ! running from 0,1,...,N/2,N/2+1,-N/2,-N/2+1,...,-1
do j = 1_pInt, grid(2)
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
k_s(2) = j - 1_pInt
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(j > grid(2)/2_pInt + 1_pInt) k_s(2) = k_s(2) - grid(2) ! running from 0,1,...,N/2,N/2+1,-N/2,-N/2+1,...,-1
do i = 1_pInt, grid1Red
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
k_s(1) = i - 1_pInt ! symmetry, junst running from 0,1,...,N/2,N/2+1
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
xi(1:3,i,j,k) = real(k_s, pReal)/scaledGeomSize ! if divergence_correction is set, frequencies are calculated on unit length
further modularisation of spectral solver (only for basic scheme so far)
2012-07-23 15:42:31 +05:30
enddo; enddo; enddo
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
if(memory_efficient) then ! allocate just single fourth order tensor
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
allocate (gamma_hat(3,3,3,3,1,1,1), source = 0.0_pReal)
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
else ! precalculation of gamma_hat field
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
allocate (gamma_hat(3,3,3,3,grid1Red ,grid(2),grid(3)), source = 0.0_pReal)
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
endif
new version of modular solver structure
2012-08-03 14:55:48 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_init
more modularization….
2012-07-26 19:28:47 +05:30
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief updates references stiffness and potentially precalculated gamma operator
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
!> @details Sets the current reference stiffness to the stiffness given as an argument.
!> If the gamma operator is precalculated, it is calculated with this stiffness.
!> In case of a on-the-fly calculation, only the reference stiffness is updated.
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!> The gamma operator is filtered depening on the filter selected in numerics.
!> Also writes out the current reference stiffness for restart.
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
subroutine utilities_updateGamma(C,saveReference)
use IO, only: &
IO_write_jobBinaryFile
new version of modular solver structure
2012-08-03 14:55:48 +05:30
use numerics, only: &
memory_efficient
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math, only: &
math_inv33
new version of modular solver structure
2012-08-03 14:55:48 +05:30
implicit none
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), intent(in), dimension(3,3,3,3) :: C !< input stiffness to store as reference stiffness
logical , intent(in) :: saveReference !< save reference stiffness to file for restart
real(pReal), dimension(3,3) :: temp33_Real, xiDyad
real(pReal) :: filter !< weighting of current component
integer(pInt) :: &
i, j, k, &
l, m, n, o
more modularization….
2012-07-26 19:28:47 +05:30
new version of modular solver structure
2012-08-03 14:55:48 +05:30
C_ref = C
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if (saveReference) then
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a)') ' writing reference stiffness to file'
flush(6)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
call IO_write_jobBinaryFile(777,'C_ref',size(C_ref))
write (777,rec=1) C_ref
close(777)
endif
new version of modular solver structure
2012-08-03 14:55:48 +05:30
if(.not. memory_efficient) then
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2); do i = 1_pInt, grid1Red
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if(any([i,j,k] /= 1_pInt)) then ! singular point at xi=(0.0,0.0,0.0) i.e. i=j=k=1
new version of modular solver structure
2012-08-03 14:55:48 +05:30
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
xiDyad(l,m) = xi(l, i,j,k)*xi(m, i,j,k)
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
fixed wrong index contraction in acoustic tensor calculation (blew up calculations with zero Poisson ratio...)
2012-11-29 00:16:07 +05:30
temp33_Real(l,m) = sum(C_ref(l,1:3,m,1:3)*xiDyad)
new version of modular solver structure
2012-08-03 14:55:48 +05:30
temp33_Real = math_inv33(temp33_Real)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
filter = utilities_getFilter(xi(1:3,i,j,k)) ! weighting factor computed by getFilter function
new version of modular solver structure
2012-08-03 14:55:48 +05:30
forall(l=1_pInt:3_pInt, m=1_pInt:3_pInt, n=1_pInt:3_pInt, o=1_pInt:3_pInt)&
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
gamma_hat(l,m,n,o, i,j,k) = filter*temp33_Real(l,n)*xiDyad(m,o)
new version of modular solver structure
2012-08-03 14:55:48 +05:30
endif
enddo; enddo; enddo
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
gamma_hat(1:3,1:3,1:3,1:3, 1,1,1) = 0.0_pReal ! singular point at xi=(0.0,0.0,0.0) i.e. i=j=k=1
new version of modular solver structure
2012-08-03 14:55:48 +05:30
endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_updateGamma
more modularization….
2012-07-26 19:28:47 +05:30
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief forward FFT of data in field_real to field_fourier with highest freqs. removed
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
!> @detailed Does an unweighted FFT transform from real to complex.
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
!> In case of debugging the FFT, also one component of the tensor (specified by row and column)
!> is independetly transformed complex to complex and compared to the whole tensor transform
new version of modular solver structure
2012-08-03 14:55:48 +05:30
!--------------------------------------------------------------------------------------------------
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
subroutine utilities_FFTforward() !< @ToDo make row and column between randomly between 1 and 3
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math
implicit none
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
integer(pInt) :: row, column ! if debug FFTW, compare 3D array field of row and column
integer(pInt), dimension(2:3,2) :: Nyquist ! highest frequencies to be removed (1 if even, 2 if odd)
real(pReal), dimension(2) :: myRand ! random numbers
next round in modularization
2012-07-25 19:31:39 +05:30
!--------------------------------------------------------------------------------------------------
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
! copy one component of the stress field to to a single FT and check for mismatch
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugFFTW) then
call random_number(myRand) ! two numbers: 0 <= x < 1
row = nint(myRand(1)*2_pReal + 1_pReal,pInt)
column = nint(myRand(2)*2_pReal + 1_pReal,pInt)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
scalarField_real = cmplx(field_real(1:grid(1),1:grid(2),1:grid(3),row,column),0.0_pReal,pReal) ! store the selected component
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
endif
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
! doing the FFT
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
call fftw_execute_dft_r2c(planForth,field_real,field_fourier)
more modularization….
2012-07-26 19:28:47 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
! comparing 1 and 3x3 FT results
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugFFTW) then
call fftw_execute_dft(planDebugForth,scalarField_real,scalarField_fourier)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
where(abs(scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3))) > tiny(1.0_pReal)) ! avoid division by zero
scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3)) = &
(scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3))-&
field_fourier(1:grid1Red,1:grid(2),1:grid(3),row,column))/&
scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3))
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
else where
scalarField_real = cmplx(0.0,0.0,pReal)
end where
write(6,'(/,a,i1,1x,i1,a)') ' .. checking FT results of compontent ', row, column, ' ..'
write(6,'(/,a,2(es11.4,1x))') ' max FT relative error = ',& ! print real and imaginary part seperately
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
maxval(real (scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3)))),&
maxval(aimag(scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3))))
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
flush(6)
endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
! removing highest frequencies
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
Nyquist(2,1:2) = [grid(2)/2_pInt + 1_pInt, grid(2)/2_pInt + 1_pInt + mod(grid(2),2_pInt)]
Nyquist(3,1:2) = [grid(3)/2_pInt + 1_pInt, grid(3)/2_pInt + 1_pInt + mod(grid(3),2_pInt)]
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(grid(1)/=1_pInt) & ! do not delete the whole slice in case of 2D calculation
field_fourier (grid1Red, 1:grid(2), 1:grid(3), 1:3,1:3) &
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
= cmplx(0.0_pReal,0.0_pReal,pReal)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(grid(2)/=1_pInt) & ! do not delete the whole slice in case of 2D calculation
field_fourier (1:grid1Red,Nyquist(2,1):Nyquist(2,2),1:grid(3), 1:3,1:3) &
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
= cmplx(0.0_pReal,0.0_pReal,pReal)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(grid(3)/=1_pInt) & ! do not delete the whole slice in case of 2D calculation
field_fourier (1:grid1Red,1:grid(2), Nyquist(3,1):Nyquist(3,2),1:3,1:3) &
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
= cmplx(0.0_pReal,0.0_pReal,pReal)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_FFTforward
more modularization….
2012-07-26 19:28:47 +05:30
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief backward FFT of data in field_fourier to field_real
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
!> @detailed Does an inverse FFT transform from complex to real
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
!> In case of debugging the FFT, also one component of the tensor (specified by row and column)
!> is independetly transformed complex to complex and compared to the whole tensor transform
!> results is weighted by number of points stored in wgt
!--------------------------------------------------------------------------------------------------
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
subroutine utilities_FFTbackward()
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math !< must use the whole module for use of FFTW
more modularization….
2012-07-26 19:28:47 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
implicit none
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
integer(pInt) :: row, column !< if debug FFTW, compare 3D array field of row and column
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
integer(pInt) :: i, j, k, m, n
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
real(pReal), dimension(2) :: myRand
more modularization….
2012-07-26 19:28:47 +05:30
!--------------------------------------------------------------------------------------------------
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
! unpack FFT data for conj complex symmetric part. This data is not transformed when using c2r
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugFFTW) then
call random_number(myRand) ! two numbers: 0 <= x < 1
row = nint(myRand(1)*2_pReal + 1_pReal,pInt)
column = nint(myRand(2)*2_pReal + 1_pReal,pInt)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
scalarField_fourier(1:grid1Red,1:grid(2),1:grid(3)) &
= field_fourier(1:grid1Red,1:grid(2),1:grid(3),row,column)
do i = 0_pInt, grid(1)/2_pInt-2_pInt + mod(grid(1),2_pInt)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
m = 1_pInt
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
n = 1_pInt
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do j = 1_pInt, grid(2)
scalarField_fourier(grid(1)-i,j,k) = conjg(scalarField_fourier(2+i,n,m))
if(n == 1_pInt) n = grid(2) + 1_pInt
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
n = n-1_pInt
enddo
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(m == 1_pInt) m = grid(3) + 1_pInt
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
m = m -1_pInt
enddo; enddo
endif
!--------------------------------------------------------------------------------------------------
! doing the iFFT
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
call fftw_execute_dft_c2r(planBack,field_fourier,field_real) ! back transform of fluct deformation gradient
new version of modular solver structure
2012-08-03 14:55:48 +05:30
more modularization….
2012-07-26 19:28:47 +05:30
!--------------------------------------------------------------------------------------------------
! comparing 1 and 3x3 inverse FT results
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugFFTW) then
call fftw_execute_dft(planDebugBack,scalarField_fourier,scalarField_real)
where(abs(real(scalarField_real,pReal)) > tiny(1.0_pReal)) ! avoid division by zero
scalarField_real = (scalarField_real &
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
- cmplx(field_real(1:grid(1),1:grid(2),1:grid(3),row,column), 0.0, pReal))/ &
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
scalarField_real
else where
scalarField_real = cmplx(0.0,0.0,pReal)
end where
write(6,'(/,a,i1,1x,i1,a)') ' ... checking iFT results of compontent ', row, column, ' ..'
write(6,'(/,a,es11.4)') ' max iFT relative error = ', maxval(real(scalarField_real,pReal))
flush(6)
endif
field_real = field_real * wgt ! normalize the result by number of elements
new version of modular solver structure
2012-08-03 14:55:48 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_FFTbackward
more modularization….
2012-07-26 19:28:47 +05:30
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!> @brief doing convolution gamma_hat * field_real, ensuring that average value = fieldAim
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
subroutine utilities_fourierConvolution(fieldAim)
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
use numerics, only: &
memory_efficient
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math, only: &
math_inv33
more modularization….
2012-07-26 19:28:47 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
implicit none
real(pReal), intent(in), dimension(3,3) :: fieldAim !< desired average value of the field after convolution
real(pReal), dimension(3,3) :: xiDyad, temp33_Real
real(pReal) :: filter !< weighting of current component
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
complex(pReal), dimension(3,3) :: temp33_complex
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
integer(pInt) :: &
i, j, k, &
l, m, n, o
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a)') ' ... doing convolution .....................................................'
flush(6)
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
introduced output of initial geometry (vtk file) to mesh
2013-04-18 22:10:49 +05:30
! do the actual spectral method calculation (mechanical equilibrium)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if(memory_efficient) then ! memory saving version, on-the-fly calculation of gamma_hat
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2) ;do i = 1_pInt, grid1Red
fixed wrong index contraction in acoustic tensor calculation (blew up calculations with zero Poisson ratio...)
2012-11-29 00:16:07 +05:30
if(any([i,j,k] /= 1_pInt)) then ! singular point at xi=(0.0,0.0,0.0) i.e. i=j=k=1
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
xiDyad(l,m) = xi(l, i,j,k)*xi(m, i,j,k)
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
temp33_Real(l,m) = sum(C_ref(l,1:3,m,1:3)*xiDyad)
temp33_Real = math_inv33(temp33_Real)
filter = utilities_getFilter(xi(1:3,i,j,k)) ! weighting factor computed by getFilter function
forall(l=1_pInt:3_pInt, m=1_pInt:3_pInt, n=1_pInt:3_pInt, o=1_pInt:3_pInt)&
gamma_hat(l,m,n,o, 1,1,1) = filter*temp33_Real(l,n)*xiDyad(m,o)
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
temp33_Complex(l,m) = sum(gamma_hat(l,m,1:3,1:3, 1,1,1) * field_fourier(i,j,k,1:3,1:3))
field_fourier(i,j,k,1:3,1:3) = temp33_Complex
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
endif
enddo; enddo; enddo
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
else ! use precalculated gamma-operator
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2); do i = 1_pInt,grid1Red
unified indices used for better understanding
2013-03-01 15:06:45 +05:30
forall(l = 1_pInt:3_pInt, m = 1_pInt:3_pInt) &
temp33_Complex(l,m) = sum(gamma_hat(l,m,1:3,1:3, i,j,k) * field_fourier(i,j,k,1:3,1:3))
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
field_fourier(i,j,k, 1:3,1:3) = temp33_Complex
enddo; enddo; enddo
endif
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
field_fourier(1,1,1,1:3,1:3) = cmplx(fieldAim*real(product(grid),pReal),0.0_pReal,pReal) ! singular point at xi=(0.0,0.0,0.0) i.e. i=j=k=1
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_fourierConvolution
more modularization….
2012-07-26 19:28:47 +05:30
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief calculate root mean square of divergence of field_fourier
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal) function utilities_divergenceRMS()
use math !< must use the whole module for use of FFTW
implicit none
integer(pInt) :: i, j, k
real(pReal) :: &
err_real_div_RMS, & !< RMS of divergence in real space
err_div_max, & !< maximum value of divergence in Fourier space
err_real_div_max !< maximum value of divergence in real space
complex(pReal), dimension(3) :: temp3_complex
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a)') ' ... calculating divergence ................................................'
flush(6)
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
more modularization….
2012-07-26 19:28:47 +05:30
! calculating RMS divergence criterion in Fourier space
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_divergenceRMS = 0.0_pReal
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2)
do i = 2_pInt, grid1Red -1_pInt ! Has somewhere a conj. complex counterpart. Therefore count it twice.
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_divergenceRMS = utilities_divergenceRMS &
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
+ 2.0_pReal*(sum (real(math_mul33x3_complex(field_fourier(i,j,k,1:3,1:3),& ! (sqrt(real(a)**2 + aimag(a)**2))**2 = real(a)**2 + aimag(a)**2. do not take square root and square again
xi(1:3,i,j,k))*TWOPIIMG)**2.0_pReal)& ! --> sum squared L_2 norm of vector
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
+sum(aimag(math_mul33x3_complex(field_fourier(i,j,k,1:3,1:3),&
xi(1:3,i,j,k))*TWOPIIMG)**2.0_pReal))
enddo
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
utilities_divergenceRMS = utilities_divergenceRMS & ! these two layers (DC and Nyquist) do not have a conjugate complex counterpart (if grid(1) /= 1)
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
+ sum( real(math_mul33x3_complex(field_fourier(1 ,j,k,1:3,1:3),&
xi(1:3,1 ,j,k))*TWOPIIMG)**2.0_pReal)&
+ sum(aimag(math_mul33x3_complex(field_fourier(1 ,j,k,1:3,1:3),&
xi(1:3,1 ,j,k))*TWOPIIMG)**2.0_pReal)&
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
+ sum( real(math_mul33x3_complex(field_fourier(grid1Red,j,k,1:3,1:3),&
xi(1:3,grid1Red,j,k))*TWOPIIMG)**2.0_pReal)&
+ sum(aimag(math_mul33x3_complex(field_fourier(grid1Red,j,k,1:3,1:3),&
xi(1:3,grid1Red,j,k))*TWOPIIMG)**2.0_pReal)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
enddo; enddo
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
if(grid(1) == 1_pInt) utilities_divergenceRMS = utilities_divergenceRMS * 0.5_pReal ! counted twice in case of grid(1) == 1
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
utilities_divergenceRMS = sqrt(utilities_divergenceRMS) * wgt ! RMS in real space calculated with Parsevals theorem from Fourier space
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
removed old standalone AL solver and introduced new structure for solver zoo
2012-07-19 22:54:56 +05:30
!--------------------------------------------------------------------------------------------------
more modularization….
2012-07-26 19:28:47 +05:30
! calculate additional divergence criteria and report
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugDivergence) then ! calculate divergence again
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
err_div_max = 0.0_pReal
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2); do i = 1_pInt, grid1Red
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
temp3_Complex = math_mul33x3_complex(field_fourier(i,j,k,1:3,1:3)*wgt,& ! weighting P_fourier
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
xi(1:3,i,j,k))*TWOPIIMG
err_div_max = max(err_div_max,sum(abs(temp3_Complex)**2.0_pReal))
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
divFourier(i,j,k,1:3) = temp3_Complex ! need divergence NOT squared
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
enddo; enddo; enddo
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
call fftw_execute_dft_c2r(planDiv,divFourier,divReal) ! already weighted
next round in modularization
2012-07-25 19:31:39 +05:30
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
err_real_div_RMS = sqrt(wgt*sum(divReal**2.0_pReal)) ! RMS in real space
err_real_div_max = sqrt(maxval(sum(divReal**2.0_pReal,dim=4))) ! max in real space
err_div_max = sqrt( err_div_max) ! max in Fourier space
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
write(6,'(/,1x,a,es11.4)') 'error divergence FT RMS = ',utilities_divergenceRMS
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
write(6,'(1x,a,es11.4)') 'error divergence Real RMS = ',err_real_div_RMS
write(6,'(1x,a,es11.4)') 'error divergence FT max = ',err_div_max
write(6,'(1x,a,es11.4)') 'error divergence Real max = ',err_real_div_max
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
flush(6)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
endif
end function utilities_divergenceRMS
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief calculate max of curl of field_fourier
!--------------------------------------------------------------------------------------------------
real(pReal) function utilities_curlRMS()
use math !< must use the whole module for use of FFTW
implicit none
integer(pInt) :: i, j, k, l
complex(pReal), dimension(3,3) :: curl_fourier
write(6,'(/,a)') ' ... calculating curl ................................................'
flush(6)
!--------------------------------------------------------------------------------------------------
! calculating max curl criterion in Fourier space
utilities_curlRMS = 0.0_pReal
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, grid(3); do j = 1_pInt, grid(2);
do i = 2_pInt, grid1Red - 1_pInt
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
do l = 1_pInt, 3_pInt
curl_fourier(l,1) = (field_fourier(i,j,k,l,3)*xi(2,i,j,k)&
- field_fourier(i,j,k,l,2)*xi(3,i,j,k))*TWOPIIMG
curl_fourier(l,2) = (-field_fourier(i,j,k,l,3)*xi(1,i,j,k)&
+field_fourier(i,j,k,l,1)*xi(3,i,j,k) )*TWOPIIMG
curl_fourier(l,3) = ( field_fourier(i,j,k,l,2)*xi(1,i,j,k)&
-field_fourier(i,j,k,l,1)*xi(2,i,j,k) )*TWOPIIMG
enddo
utilities_curlRMS = utilities_curlRMS + &
2.0_pReal*sum(real(curl_fourier)**2.0_pReal + aimag(curl_fourier)**2.0_pReal)
enddo
do l = 1_pInt, 3_pInt
curl_fourier = (field_fourier(1,j,k,l,3)*xi(2,1,j,k)&
- field_fourier(1,j,k,l,2)*xi(3,1,j,k))*TWOPIIMG
curl_fourier = (-field_fourier(1,j,k,l,3)*xi(1,1,j,k)&
+field_fourier(1,j,k,l,1)*xi(3,1,j,k) )*TWOPIIMG
curl_fourier = ( field_fourier(1,j,k,l,2)*xi(1,1,j,k)&
-field_fourier(1,j,k,l,1)*xi(2,1,j,k) )*TWOPIIMG
enddo
utilities_curlRMS = utilities_curlRMS + &
2.0_pReal*sum(real(curl_fourier)**2.0_pReal + aimag(curl_fourier)**2.0_pReal)
do l = 1_pInt, 3_pInt
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
curl_fourier = ( field_fourier(grid1Red,j,k,l,3)*xi(2,grid1Red,j,k)&
-field_fourier(grid1Red,j,k,l,2)*xi(3,grid1Red,j,k))*TWOPIIMG
curl_fourier = (-field_fourier(grid1Red,j,k,l,3)*xi(1,grid1Red,j,k)&
+field_fourier(grid1Red,j,k,l,1)*xi(3,grid1Red,j,k))*TWOPIIMG
curl_fourier = ( field_fourier(grid1Red,j,k,l,2)*xi(1,grid1Red,j,k)&
-field_fourier(grid1Red,j,k,l,1)*xi(2,grid1Red,j,k))*TWOPIIMG
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
enddo
utilities_curlRMS = utilities_curlRMS + &
2.0_pReal*sum(real(curl_fourier)**2.0_pReal + aimag(curl_fourier)**2.0_pReal)
enddo; enddo
utilities_curlRMS = sqrt(utilities_curlRMS) *wgt
end function utilities_curlRMS
next round in modularizzation of solvers
2012-07-24 22:37:10 +05:30
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
introduced output of initial geometry (vtk file) to mesh
2013-04-18 22:10:49 +05:30
!> @brief calculates mask compliance tensor used to adjust F to fullfill stress BC
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
!--------------------------------------------------------------------------------------------------
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
function utilities_maskedCompliance(rot_BC,mask_stress,C)
use IO, only: &
IO_error
use math, only: &
math_Plain3333to99, &
math_plain99to3333, &
math_rotate_forward3333, &
math_rotate_forward33, &
math_invert
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
implicit none
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), dimension(3,3,3,3) :: utilities_maskedCompliance !< masked compliance
real(pReal), intent(in) , dimension(3,3,3,3) :: C !< current average stiffness
real(pReal), intent(in) , dimension(3,3) :: rot_BC !< rotation of load frame
logical, intent(in), dimension(3,3) :: mask_stress !< mask of stress BC
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
integer(pInt) :: j, k, m, n
unified naming scheme, fixed a bug in the new basic scheme, and added a statistic file similar to abaqus to store the information on needed cut backs and iterations for each inc
2012-10-19 14:14:21 +05:30
logical, dimension(9) :: mask_stressVector
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
real(pReal), dimension(9,9) :: temp99_Real
integer(pInt) :: size_reduced = 0_pInt
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), dimension(:,:), allocatable :: &
s_reduced, & !< reduced compliance matrix (depending on number of stress BC)
c_reduced, & !< reduced stiffness (depending on number of stress BC)
sTimesC !< temp variable to check inversion
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
logical :: errmatinv
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
character(len=1024):: formatString
next round in modularization
2012-07-25 19:31:39 +05:30
unified naming scheme, fixed a bug in the new basic scheme, and added a statistic file similar to abaqus to store the information on needed cut backs and iterations for each inc
2012-10-19 14:14:21 +05:30
mask_stressVector = reshape(transpose(mask_stress), [9])
size_reduced = int(count(mask_stressVector), pInt)
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
if(size_reduced > 0_pInt )then
allocate (c_reduced(size_reduced,size_reduced), source =0.0_pReal)
allocate (s_reduced(size_reduced,size_reduced), source =0.0_pReal)
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
allocate (sTimesC(size_reduced,size_reduced), source =0.0_pReal)
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
temp99_Real = math_Plain3333to99(math_rotate_forward3333(C,rot_BC))
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
if(debugGeneral) then
write(6,'(/,a)') ' ... updating masked compliance ............................................'
introduced output of initial geometry (vtk file) to mesh
2013-04-18 22:10:49 +05:30
write(6,'(/,a,/,9(9(2x,f12.7,1x)/))',advance='no') ' Stiffness C (load) / GPa =',&
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
transpose(temp99_Real)/1.e9_pReal
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
flush(6)
endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
k = 0_pInt ! calculate reduced stiffness
more modularization….
2012-07-26 19:28:47 +05:30
do n = 1_pInt,9_pInt
if(mask_stressVector(n)) then
k = k + 1_pInt
j = 0_pInt
do m = 1_pInt,9_pInt
if(mask_stressVector(m)) then
j = j + 1_pInt
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
c_reduced(k,j) = temp99_Real(n,m)
endif; enddo; endif; enddo
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
call math_invert(size_reduced, c_reduced, s_reduced, errmatinv) ! invert reduced stiffness
made restart working for all solvers and added corresponding test corrected error handling for FFT based function in case of odd numbers etc.
2012-12-14 23:00:22 +05:30
if(errmatinv) call IO_error(error_ID=400_pInt,ext_msg='utilities_maskedCompliance')
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
temp99_Real = 0.0_pReal ! fill up compliance with zeros
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
k = 0_pInt
do n = 1_pInt,9_pInt
if(mask_stressVector(n)) then
k = k + 1_pInt
j = 0_pInt
do m = 1_pInt,9_pInt
if(mask_stressVector(m)) then
j = j + 1_pInt
temp99_Real(n,m) = s_reduced(k,j)
endif; enddo; endif; enddo
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
! check if inversion was successful
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
sTimesC = matmul(c_reduced,s_reduced)
do m=1_pInt, size_reduced
do n=1_pInt, size_reduced
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if(m==n .and. abs(sTimesC(m,n)) > (1.0_pReal + 10.0e-12_pReal)) errmatinv = .true. ! diagonal elements of S*C should be 1
if(m/=n .and. abs(sTimesC(m,n)) > (0.0_pReal + 10.0e-12_pReal)) errmatinv = .true. ! off diagonal elements of S*C should be 0
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
enddo
enddo
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if(debugGeneral .or. errmatinv) then ! report
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
write(formatString, '(I16.16)') size_reduced
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
formatString = '(/,a,/,'//trim(formatString)//'('//trim(formatString)//'(2x,es9.2,1x)/))'
introduced output of initial geometry (vtk file) to mesh
2013-04-18 22:10:49 +05:30
write(6,trim(formatString),advance='no') ' C * S (load) ', &
transpose(matmul(c_reduced,s_reduced))
write(6,trim(formatString),advance='no') ' S (load) ', transpose(s_reduced)
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
endif
made restart working for all solvers and added corresponding test corrected error handling for FFT based function in case of odd numbers etc.
2012-12-14 23:00:22 +05:30
if(errmatinv) call IO_error(error_ID=400_pInt,ext_msg='utilities_maskedCompliance')
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
deallocate(c_reduced)
deallocate(s_reduced)
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
deallocate(sTimesC)
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
else
temp99_real = 0.0_pReal
endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if(debugGeneral) & ! report
introduced output of initial geometry (vtk file) to mesh
2013-04-18 22:10:49 +05:30
write(6,'(/,a,/,9(9(2x,f12.7,1x)/),/)',advance='no') ' Masked Compliance (load) * GPa =', &
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
transpose(temp99_Real*1.e9_pReal)
flush(6)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_maskedCompliance = math_Plain99to3333(temp99_Real)
end function utilities_maskedCompliance
!--------------------------------------------------------------------------------------------------
!> @brief calculates constitutive response
!--------------------------------------------------------------------------------------------------
removed coordinates from call to CPFEM_general and introduced direct storage of coordinates to mesh_ipCoordinates
2012-11-14 20:08:10 +05:30
subroutine utilities_constitutiveResponse(F_lastInc,F,temperature,timeinc,&
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
P,C_volAvg,C_minmaxAvg,P_av,forwardData,rotation_BC)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use debug, only: &
debug_reset, &
debug_info
use math, only: &
math_transpose33, &
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
math_rotate_forward33, &
math_det33
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use FEsolving, only: &
restartWrite
use CPFEM, only: &
reorganized calculation modes for CPFEM, now having better readable and cleaner structure
2013-03-01 17:18:29 +05:30
CPFEM_general, &
CPFEM_COLLECT, &
CPFEM_CALCRESULTS, &
CPFEM_AGERESULTS, &
CPFEM_BACKUPJACOBIAN, &
CPFEM_RESTOREJACOBIAN
checked and improved on Franz' one call to CPFEM only improvement
2013-01-24 01:26:45 +05:30
use homogenization, only: &
materialpoint_F0, &
materialpoint_F, &
materialpoint_Temperature, &
materialpoint_P, &
materialpoint_dPdF
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
implicit none
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
real(pReal), intent(inout) :: temperature !< temperature (no field)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), intent(in), dimension(3,3,grid(1),grid(2),grid(3)) :: &
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
F_lastInc, & !< target deformation gradient
F !< previous deformation gradient
real(pReal), intent(in) :: timeinc !< loading time
logical, intent(in) :: forwardData !< age results
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), intent(in), dimension(3,3) :: rotation_BC !< rotation of load frame
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal),intent(out), dimension(3,3,3,3) :: C_volAvg, C_minmaxAvg !< average stiffness
real(pReal),intent(out), dimension(3,3) :: P_av !< average PK stress
real(pReal),intent(out), dimension(3,3,grid(1),grid(2),grid(3)) :: P !< PK stress
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
integer(pInt) :: &
calcMode, & !< CPFEM mode for calculation
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
collectMode, & !< CPFEM mode for collection
j,k
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
real(pReal), dimension(3,3,3,3) :: max_dPdF, min_dPdF
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
real(pReal) :: max_dPdF_norm, min_dPdF_norm, defgradDetMin, defgradDetMax, defgradDet
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a)') ' ... evaluating constitutive response ......................................'
reorganized calculation modes for CPFEM, now having better readable and cleaner structure
2013-03-01 17:18:29 +05:30
calcMode = CPFEM_CALCRESULTS
collectMode = CPFEM_COLLECT
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
if (forwardData) then ! aging results
reorganized calculation modes for CPFEM, now having better readable and cleaner structure
2013-03-01 17:18:29 +05:30
calcMode = ior(calcMode, CPFEM_AGERESULTS)
collectMode = ior(collectMode, CPFEM_BACKUPJACOBIAN)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
endif
if (cutBack) then ! restore saved variables
reorganized calculation modes for CPFEM, now having better readable and cleaner structure
2013-03-01 17:18:29 +05:30
collectMode = ior(collectMode , CPFEM_RESTOREJACOBIAN)
collectMode = iand(collectMode, not(CPFEM_BACKUPJACOBIAN))
calcMode = iand(calcMode, not(CPFEM_AGERESULTS))
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
endif
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
call CPFEM_general(collectMode,F_lastInc(1:3,1:3,1,1,1),F(1:3,1:3,1,1,1), & ! collect mode handles Jacobian backup / restoration
temperature,timeinc,1_pInt,1_pInt)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
materialpoint_F0 = reshape(F_lastInc, [3,3,1,product(grid)])
materialpoint_F = reshape(F, [3,3,1,product(grid)])
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
materialpoint_Temperature = temperature
call debug_reset()
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
!--------------------------------------------------------------------------------------------------
! calculate bounds of det(F) and report
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
if(debugGeneral) then
defgradDetMax = -huge(1.0_pReal)
defgradDetMin = +huge(1.0_pReal)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do j = 1_pInt, product(grid)
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
defgradDet = math_det33(materialpoint_F(1:3,1:3,1,j))
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
defgradDetMax = max(defgradDetMax,defgradDet)
defgradDetMin = min(defgradDetMin,defgradDet)
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
end do
simplified CurlFFT and DivergenceFFT functions, and did the last changes to Utilities related to the new structure of the spectral solver
2013-03-07 01:04:30 +05:30
write(6,'(a,1x,es11.4)') ' max determinant of deformation =', defgradDetMax
write(6,'(a,1x,es11.4)') ' min determinant of deformation =', defgradDetMin
flush(6)
endif
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
checked and improved on Franz' one call to CPFEM only improvement
2013-01-24 01:26:45 +05:30
call CPFEM_general(calcMode,F_lastInc(1:3,1:3,1,1,1), F(1:3,1:3,1,1,1), & ! first call calculates everything
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
temperature,timeinc,1_pInt,1_pInt)
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
max_dPdF = 0.0_pReal
max_dPdF_norm = 0.0_pReal
min_dPdF = huge(1.0_pReal)
min_dPdF_norm = huge(1.0_pReal)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
do k = 1_pInt, product(grid)
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
if (max_dPdF_norm < sum(materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)**2.0_pReal)) then
max_dPdF = materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)
max_dPdF_norm = sum(materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)**2.0_pReal)
endif
if (min_dPdF_norm > sum(materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)**2.0_pReal)) then
min_dPdF = materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)
min_dPdF_norm = sum(materialpoint_dPdF(1:3,1:3,1:3,1:3,1,k)**2.0_pReal)
endif
simplified interface of CPFEM_general (removed P and F, made cs and dcs/dE optional) commented and cleaned up the marc interface. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! For marc simulations, run ./code/setup/setup_code.sh !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2013-04-09 15:38:00 +05:30
end do
checked and improved on Franz' one call to CPFEM only improvement
2013-01-24 01:26:45 +05:30
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
P = reshape(materialpoint_P, [3,3,grid(1),grid(2),grid(3)])
changed reference stiffness to min max avg from volume avg. using terminally ill information in convergence check
2013-03-06 20:01:13 +05:30
C_volAvg = sum(sum(materialpoint_dPdF,dim=6),dim=5) * wgt
C_minmaxAvg = 0.5_pReal*(max_dPdF + min_dPdF)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
call debug_info()
restartWrite = .false. ! reset restartWrite status
cutBack = .false. ! reset cutBack status
simplified the rate calculation interface in DAMASK_spectral_utilities.f90, corrected bug in rotation parsing and added debug option "rotation" to spectral solver options. This will show the current average stress and deformation additionally in lab coordinate system
2012-11-28 20:34:05 +05:30
P_av = sum(sum(sum(P,dim=5),dim=4),dim=3) * wgt ! average of P
if (debugRotation) &
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a,/,3(3(2x,f12.7,1x)/))',advance='no') ' Piola--Kirchhoff stress (lab) / MPa =',&
simplified the rate calculation interface in DAMASK_spectral_utilities.f90, corrected bug in rotation parsing and added debug option "rotation" to spectral solver options. This will show the current average stress and deformation additionally in lab coordinate system
2012-11-28 20:34:05 +05:30
math_transpose33(P_av)/1.e6_pReal
P_av = math_rotate_forward33(P_av,rotation_BC)
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
write(6,'(/,a,/,3(3(2x,f12.7,1x)/))',advance='no') ' Piola--Kirchhoff stress / MPa =',&
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
math_transpose33(P_av)/1.e6_pReal
corrected reporting of basic PETSc variant and unified reporting of all spectral solvers improved warning and error in IO, now able to report correctly elements up to 9 digits
2013-01-10 03:49:32 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_constitutiveResponse
further work on modularization, basic scheme is now calculating the same results as DAMASK_spectral.f90
2012-07-30 19:36:22 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief calculates forward rate, either guessing or just add delta/timeinc
!--------------------------------------------------------------------------------------------------
simplified the rate calculation interface in DAMASK_spectral_utilities.f90, corrected bug in rotation parsing and added debug option "rotation" to spectral solver options. This will show the current average stress and deformation additionally in lab coordinate system
2012-11-28 20:34:05 +05:30
pure function utilities_calculateRate(avRate,timeinc_old,guess,field_lastInc,field)
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
substituted hand written matrix inversion by LAPACK version with precision selection. also introduced check for inversion into DAMASK_spectral_Utilities.f90 for the stress BC calculation. This part is further improved by using 5% of the reference stiffness to avoid trouble in the fully plastic regime (where the stiffness is underestimated) Test for Marc 2010 is updated because the new inversion give slightly different results near 0 (order of e-13)
2012-08-28 22:29:45 +05:30
implicit none
simplified the rate calculation interface in DAMASK_spectral_utilities.f90, corrected bug in rotation parsing and added debug option "rotation" to spectral solver options. This will show the current average stress and deformation additionally in lab coordinate system
2012-11-28 20:34:05 +05:30
real(pReal), intent(in), dimension(3,3) :: avRate !< homogeneous addon
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
real(pReal), intent(in) :: &
replaced real 'guessmode' with logical 'guess', caused trouble on gfortran 6.3 with ultra optimization
2012-11-09 01:02:00 +05:30
timeinc_old !< timeinc of last step
logical, intent(in) :: &
guess !< guess along former trajectory
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), intent(in), dimension(3,3,grid(1),grid(2),grid(3)) :: &
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
field_lastInc, & !< data of previous step
field !< data of current step
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), dimension(3,3,grid(1),grid(2),grid(3)) :: utilities_calculateRate
new version of modular solver structure
2012-08-03 14:55:48 +05:30
added comments and structured the code. Temperature is not longer stored for each point but only to simplify restart behavior as long as it is not fully supported by the constitutive models
2012-11-12 19:44:39 +05:30
if(guess) then
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_calculateRate = (field-field_lastInc) / timeinc_old
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
else
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
utilities_calculateRate = spread(spread(spread(avRate,3,grid(1)),4,grid(2)),5,grid(3))
changed order of most arrays to fortran-fast, whole new solver still work in progress
2012-08-09 18:34:56 +05:30
endif
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end function utilities_calculateRate
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
!> @brief forwards a field with a pointwise given rate, if aim is given,
!> ensures that the average matches the aim
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
pure function utilities_forwardField(timeinc,field_lastInc,rate,aim)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
implicit none
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
real(pReal), intent(in) :: &
timeinc !< timeinc of current step
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), intent(in), dimension(3,3,grid(1),grid(2),grid(3)) :: &
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
field_lastInc, & !< initial field
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
rate !< rate by which to forward
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
real(pReal), intent(in), optional, dimension(3,3) :: &
aim !< average field value aim
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
real(pReal), dimension(3,3,grid(1),grid(2),grid(3)) :: utilities_forwardField
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
real(pReal), dimension(3,3) :: fieldDiff !< <a + adot*t> - aim
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
utilities_forwardField = field_lastInc + rate*timeinc
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
if (present(aim)) then !< correct to match average
fieldDiff = sum(sum(sum(utilities_forwardField,dim=5),dim=4),dim=3)*wgt - aim
utilities_forwardField = utilities_forwardField - &
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
spread(spread(spread(fieldDiff,3,grid(1)),4,grid(2)),5,grid(3))
fixed bug in forwarding fields for AL solver
2013-01-08 15:42:03 +05:30
endif
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end function utilities_forwardField
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief calculates filter for fourier convolution depending on type given in numerics.config
!--------------------------------------------------------------------------------------------------
real(pReal) function utilities_getFilter(k)
use IO, only: &
IO_error
use numerics, only: &
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
myfilter
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
use math, only: &
PI
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
implicit none
real(pReal),intent(in), dimension(3) :: k !< indices of frequency
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
fixed bug in coordinate calculation when no average F was given (scaled wrong by ncp_elems**2) other changes: just polishing + some more comments
2013-03-27 17:58:55 +05:30
utilities_getFilter = 1.0_pReal
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
select case (myfilter)
added missing line continuation in numerics and explicit private statements in the other files
2013-03-28 16:07:00 +05:30
case ('none') !< default is already nothing (1.0_pReal)
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
case ('cosine') !< cosine curve with 1 for avg and zero for highest freq
moved public data res,size and homog from mesh to DAMASK_spectral_utilities (as grid and geomSize)
2013-05-08 21:22:29 +05:30
utilities_getFilter = (1.0_pReal + cos(PI*k(3)/grid(3))) &
*(1.0_pReal + cos(PI*k(2)/grid(2))) &
*(1.0_pReal + cos(PI*k(1)/grid(1)))/8.0_pReal
new solver is now compiling without a PETSc installation, however only the plain basic solver is available then. also removed reporting of PETSc related variables in the case it is not installed/found
2012-08-29 00:49:47 +05:30
case default
call IO_error(error_ID = 892_pInt, ext_msg = trim(myfilter))
added filter option to the spectral solver to mitigate spurious oscillations due to gibb's phenomenon. activate by setting myfilter in config file appropriately (currently only 'none' and 'cosine' options coded). more cleaning up of AL code
2012-08-07 22:53:13 +05:30
end select
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end function utilities_getFilter
new version of modular solver structure
2012-08-03 14:55:48 +05:30
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
!--------------------------------------------------------------------------------------------------
!> @brief cleans up
!--------------------------------------------------------------------------------------------------
subroutine utilities_destroy()
use math
added missing line continuation in numerics and explicit private statements in the other files
2013-03-28 16:07:00 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
implicit none
next round in modularization
2012-07-25 19:31:39 +05:30
allowed arbitrary grid (formerly know as resolution), i.e. any integer>1 in each dimension is possible. removed square root correction for divergence (doesn't make sense) fixed FFT debug of spectral solver
2013-04-10 15:49:16 +05:30
if (debugDivergence) call fftw_destroy_plan(planDiv)
if (debugFFTW) call fftw_destroy_plan(planDebugForth)
if (debugFFTW) call fftw_destroy_plan(planDebugBack)
call fftw_destroy_plan(planForth)
call fftw_destroy_plan(planBack)
next round in modularization
2012-07-25 19:31:39 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end subroutine utilities_destroy
numerics got some protected statements and is now reading in the new keywords for the solver selection in small letters. spectral solver got cut back facilities + improved output to screen
2012-10-02 20:56:56 +05:30
documented utilities and structured, worked on the restart capabilities of the new basic solver
2012-10-24 17:01:40 +05:30
end module DAMASK_spectral_utilities