From aeae15a3abf4c930125c0088f21cfedc117c91a1 Mon Sep 17 00:00:00 2001
From: Jean-Matthieu Etancelin <jean-matthieu.etancelin@univ-reims.fr>
Date: Tue, 28 Oct 2014 10:46:24 +0100
Subject: [PATCH] Ajout patch pour passer en simple precision. Corrections de
tests pour la simple presision.
---
HySoP/hysop/domain/tests/test_control_box.py | 2 +
HySoP/hysop/domain/tests/test_mesh.py | 6 +-
.../hysop/domain/tests/test_regular_subset.py | 3 +
HySoP/hysop/domain/tests/test_subset.py | 12 +-
ToSinglePrecision.patch | 952 ++++++++++++++++++
5 files changed, 967 insertions(+), 8 deletions(-)
create mode 100644 ToSinglePrecision.patch
diff --git a/HySoP/hysop/domain/tests/test_control_box.py b/HySoP/hysop/domain/tests/test_control_box.py
index 20d85b710..dc7b2cda2 100644
--- a/HySoP/hysop/domain/tests/test_control_box.py
+++ b/HySoP/hysop/domain/tests/test_control_box.py
@@ -42,6 +42,8 @@ xdef = orig + 0.2
def check_control_box(discr, xr, lr):
+ xr = npw.asrealarray(xr)
+ lr = npw.asrealarray(lr)
dim = len(discr.resolution)
dom = Box(dimension=dim, length=ldom[:dim],
origin=orig[:dim])
diff --git a/HySoP/hysop/domain/tests/test_mesh.py b/HySoP/hysop/domain/tests/test_mesh.py
index 0999bd537..5cdb6a168 100644
--- a/HySoP/hysop/domain/tests/test_mesh.py
+++ b/HySoP/hysop/domain/tests/test_mesh.py
@@ -5,7 +5,7 @@ Testing regular grids.
from parmepy.domain.box import Box
from parmepy.tools.parameters import Discretization
from parmepy.mpi import main_size, main_rank
-import numpy as np
+import parmepy.tools.numpywrappers as npw
Nx = Ny = Nz = 32
@@ -26,7 +26,7 @@ def create_mesh(discr):
assert mesh.discretization == discr
assert (mesh.space_step == dom.length / (discr.resolution - 1)).all()
# Position compared with global grid
- shift = np.asarray([0, ] * dim)
+ shift = npw.asrealarray([0, ] * dim)
shift[-1] = Nz / main_size * main_rank
resolution = (discr.resolution - 1) / topo.shape + 2 * gh
end = shift + resolution - 2 * gh
@@ -52,7 +52,7 @@ def create_mesh(discr):
assert (mesh.global_indices(point) == req_point).all()
if main_rank == main_size - 1:
assert mesh.is_inside(point)
- pos = np.asarray(req_point)
+ pos = npw.asrealarray(req_point)
pos[-1] = Nz / main_size - 1
pos += gh
assert (mesh.local_indices(point) == pos).all()
diff --git a/HySoP/hysop/domain/tests/test_regular_subset.py b/HySoP/hysop/domain/tests/test_regular_subset.py
index d2d36b2a0..1c1fd4dc4 100644
--- a/HySoP/hysop/domain/tests/test_regular_subset.py
+++ b/HySoP/hysop/domain/tests/test_regular_subset.py
@@ -1,6 +1,7 @@
from parmepy.domain.subsets.boxes import SubBox
from parmepy.tools.parameters import Discretization
from parmepy import Field, Box
+import parmepy.tools.numpywrappers as npw
import numpy as np
import math
@@ -40,6 +41,8 @@ xdef = xdom + 0.2
def check_subset(discr, xr, lr):
+ xr = npw.asrealarray(xr)
+ lr = npw.asrealarray(lr)
dim = len(discr.resolution)
dom = Box(dimension=dim, length=ldom[:dim],
origin=xdom[:dim])
diff --git a/HySoP/hysop/domain/tests/test_subset.py b/HySoP/hysop/domain/tests/test_subset.py
index 555916e67..9c5bb9498 100644
--- a/HySoP/hysop/domain/tests/test_subset.py
+++ b/HySoP/hysop/domain/tests/test_subset.py
@@ -6,9 +6,11 @@ from parmepy.operator.hdf_io import HDF_Reader
from parmepy.tools.parameters import Discretization, IO_params
from parmepy import Field, Box
from parmepy.mpi.topology import Cartesian
+import parmepy.tools.numpywrappers as npw
import numpy as np
import math
from parmepy.mpi import main_size
+import mpi4py
Nx = 128
@@ -17,13 +19,13 @@ Nz = 102
g = 2
-ldef = [0.3, 0.4, 1.0]
+ldef = npw.asrealarray([0.3, 0.4, 1.0])
discr3D = Discretization([Nx + 1, Ny + 1, Nz + 1], [g - 1, g - 2, g])
discr2D = Discretization([Nx + 1, Ny + 1], [g - 1, g])
-xdom = np.asarray([0.1, -0.3, 0.5])
-ldom = np.asarray([math.pi * 2., ] * 3)
-xdef = xdom + 0.2
-xpos = ldom * 0.5
+xdom = npw.asrealarray([0.1, -0.3, 0.5])
+ldom = npw.asrealarray([math.pi * 2., ] * 3)
+xdef = npw.asrealarray(xdom + 0.2)
+xpos = npw.asrealarray(ldom * 0.5)
xpos[-1] += 0.1
import os
working_dir = os.getcwd() + '/p' + str(main_size) + '/'
diff --git a/ToSinglePrecision.patch b/ToSinglePrecision.patch
new file mode 100644
index 000000000..4632059ed
--- /dev/null
+++ b/ToSinglePrecision.patch
@@ -0,0 +1,952 @@
+diff --git parmepy/constants.py parmepy/constants.py
+index d77a3e9..ee24b47 100644
+--- parmepy/constants.py
++++ parmepy/constants.py
+@@ -10,7 +10,7 @@ from parmepy.mpi import MPI
+
+ PI = math.pi
+ # Set default type for real and integer numbers
+-PARMES_REAL = np.float64
++PARMES_REAL = np.float32
+ SIZEOF_PARMES_REAL = int(PARMES_REAL(1.).nbytes)
+ # type for array indices
+ PARMES_INDEX = np.uint32
+@@ -19,7 +19,7 @@ PARMES_INTEGER = np.int32
+ # integer used for arrays dimensions
+ PARMES_DIM = np.int16
+ # float type for MPI messages
+-PARMES_MPI_REAL = MPI.DOUBLE
++PARMES_MPI_REAL = MPI.REAL
+ # int type for MPI messages
+ PARMES_MPI_INTEGER = MPI.INT
+ ## default array layout (fortran or C convention)
+diff --git parmepy/f2py/fftw2py.f90 parmepy/f2py/fftw2py.f90
+index c032cd0..113db90 100755
+--- parmepy/f2py/fftw2py.f90
++++ parmepy/f2py/fftw2py.f90
+@@ -5,7 +5,7 @@
+ module fftw2py
+
+ use client_data
+- use parmesparam
++ use parmesparam_sp
+ !> 2d case
+ use fft2d
+ !> 3d case
+@@ -81,7 +81,7 @@ contains
+ subroutine solve_poisson_2d(omega,velocity_x,velocity_y)
+ real(pk),dimension(:,:),intent(in):: omega
+ real(pk),dimension(size(omega,1),size(omega,2)),intent(out) :: velocity_x,velocity_y
+- real(pk) :: start
++ real(8) :: start
+ !f2py intent(in,out) :: velocity_x,velocity_y
+ start = MPI_WTime()
+
+@@ -117,7 +117,7 @@ contains
+ real(pk),dimension(:,:,:),intent(in):: omega_x,omega_y,omega_z
+ real(pk),dimension(size(omega_x,1),size(omega_y,2),size(omega_z,3)),intent(out) :: velocity_x,velocity_y,velocity_z
+ integer, dimension(3), intent(in) :: ghosts_vort, ghosts_velo
+- real(pk) :: start
++ real(8) :: start
+ !f2py intent(in,out) :: velocity_x,velocity_y,velocity_z
+ start = MPI_WTime()
+ call r2c_3d(omega_x,omega_y,omega_z, ghosts_vort)
+diff --git parmepy/f2py/scales2py.f90 parmepy/f2py/scales2py.f90
+index 7884535..f901685 100755
+--- parmepy/f2py/scales2py.f90
++++ parmepy/f2py/scales2py.f90
+@@ -6,7 +6,7 @@ use advec, only : advec_init,advec_step,advec_step_Inter_basic,advec_step_Inter_
+ use advec_vect, only : advec_step_Vect,advec_step_Inter_basic_Vect
+ use interpolation_velo, only : interpol_init
+ use mpi
+-use parmesparam
++use parmesparam_sp
+
+
+ implicit none
+@@ -93,7 +93,7 @@ contains
+ real(pk), dimension(size(vx,1),size(vx,2),size(vx,3)), intent(inout) :: scal
+ !f2py real(pk) intent(in,out), depend(size(vx,1)) :: scal
+
+- real(pk) :: t0
++ real(8) :: t0
+
+ t0 = MPI_Wtime()
+ call advec_step(dt,vx,vy,vz,scal)
+diff --git setup.py.in setup.py.in
+index dec8ebd..559ad8c 100644
+--- setup.py.in
++++ setup.py.in
+@@ -70,8 +70,8 @@ if enable_fortran is "ON":
+ fortran_src.append(fortran_dir+'parameters.f90')
+ fortran_src.append(fortran_dir+'fftw2py.f90')
+ fftwdir = '@FFTWLIB@'
+- parmeslib.append('fftw3')
+- parmeslib.append('fftw3_mpi')
++ parmeslib.append('fftw3f')
++ parmeslib.append('fftw3f_mpi')
+ parmes_libdir.append(fftwdir)
+ else:
+ packages.append('parmepy.fakef2py')
+diff --git src/client_data.f90 src/client_data.f90
+index 46b5268..77178d9 100755
+--- src/client_data.f90
++++ src/client_data.f90
+@@ -1,14 +1,14 @@
+ !> Some global parameters and variables
+ module client_data
+
+- use MPI, only : MPI_DOUBLE_PRECISION
++ use MPI, only : MPI_REAL
+ use, intrinsic :: iso_c_binding ! required for fftw
+ implicit none
+
+ !> kind for real variables (simple or double precision)
+- integer, parameter :: mk = kind(1.0d0) ! double precision
++ integer, parameter :: mk = kind(1.0) ! single precision
+ !> kind for real variables in mpi routines
+- integer, parameter :: mpi_mk = MPI_DOUBLE_PRECISION
++ integer, parameter :: mpi_mk = MPI_REAL
+ !> Problem dimension (model, required for ppm to work properly)
+ integer, parameter :: dime = 2
+ !> Real dimension
+@@ -26,7 +26,7 @@ module client_data
+ !> to activate (or not) screen output
+ logical,parameter :: verbose = .True.
+ !> i (sqrt(-1) ...)
+- complex(C_DOUBLE_COMPLEX), parameter :: Icmplx = cmplx(0._mk,1._mk, kind=mk)
++ complex(C_FLOAT_COMPLEX), parameter :: Icmplx = cmplx(0._mk,1._mk, kind=mk)
+ !> tolerance used to compute error
+ real(mk), parameter :: tolerance = 1e-12
+
+diff --git src/fftw/Poisson.f90 src/fftw/Poisson.f90
+index 78b355c..8804e3d 100755
+--- src/fftw/Poisson.f90
++++ src/fftw/Poisson.f90
+@@ -53,7 +53,7 @@ contains
+ real(mk),dimension(:,:,:),intent(in) :: omega_x,omega_y,omega_z
+ real(mk),dimension(:,:,:),intent(inout) :: velocity_x,velocity_y,velocity_z
+ integer, dimension(3), intent(in) :: ghosts_w, ghosts_v
+- real(mk) :: start
++ real(8) :: start
+ !! Compute fftw forward transform
+ !! Omega is used to initialize the fftw buffer for input field.
+
+@@ -74,7 +74,7 @@ contains
+
+ real(mk),dimension(:,:,:),intent(in) :: omega_x,omega_y,omega_z
+ real(mk),dimension(:,:,:),intent(inout) :: velocity_x,velocity_y,velocity_z
+- real(mk) :: start
++ real(8) :: start
+ !! Compute fftw forward transform
+ !! Omega is used to initialize the fftw buffer for input field.
+
+diff --git src/fftw/fft2d.f90 src/fftw/fft2d.f90
+index cf2a69a..fd0b337 100755
+--- src/fftw/fft2d.f90
++++ src/fftw/fft2d.f90
+@@ -42,15 +42,15 @@ module fft2d
+ type(C_PTR) :: cbuffer2
+ !! Note Franck : check if local declarations of datain/out works and improve perfs.
+ !> Field (complex values) for fftw input
+- complex(C_DOUBLE_COMPLEX), pointer :: datain1(:,:),datain2(:,:)
++ complex(C_FLOAT_COMPLEX), pointer :: datain1(:,:),datain2(:,:)
+ !> Field (real values) for fftw input
+- real(C_DOUBLE), pointer :: rdatain1(:,:)
++ real(C_FLOAT), pointer :: rdatain1(:,:)
+ !> Field (complex values) for fftw (forward) output
+- complex(C_DOUBLE_COMPLEX), pointer :: dataout1(:,:)
++ complex(C_FLOAT_COMPLEX), pointer :: dataout1(:,:)
+ !> Field (real values) for fftw output
+- real(C_DOUBLE), pointer :: rdatain2(:,:)
++ real(C_FLOAT), pointer :: rdatain2(:,:)
+ !> Field (complex values) for fftw (forward) output
+- complex(C_DOUBLE_COMPLEX), pointer :: dataout2(:,:)
++ complex(C_FLOAT_COMPLEX), pointer :: dataout2(:,:)
+ !> GLOBAL number of points in each direction
+ integer(C_INTPTR_T),pointer :: fft_resolution(:)
+ !> LOCAL resolution
+@@ -58,13 +58,13 @@ module fft2d
+ !> Offset in the direction of distribution
+ integer(c_INTPTR_T),dimension(2) :: local_offset
+ !> wave numbers for fft in x direction
+- real(C_DOUBLE), pointer :: kx(:)
++ real(C_FLOAT), pointer :: kx(:)
+ !> wave numbers for fft in y direction
+- real(C_DOUBLE), pointer :: ky(:)
++ real(C_FLOAT), pointer :: ky(:)
+ !> log file for fftw
+ character(len=20),parameter :: filename ="parmesfftw.log"
+ !> normalization factor
+- real(C_DOUBLE) :: normFFT
++ real(C_FLOAT) :: normFFT
+ !> true if all the allocation stuff for global variables has been done.
+ logical :: is2DUpToDate = .false.
+
+@@ -88,7 +88,7 @@ contains
+ if(is2DUpToDate) return
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+
+@@ -98,13 +98,13 @@ contains
+
+ ! compute "optimal" size (according to fftw) for local date
+ ! (warning : dimension reversal)
+- alloc_local = fftw_mpi_local_size_2d_transposed(fft_resolution(c_Y), &
++ alloc_local = fftwf_mpi_local_size_2d_transposed(fft_resolution(c_Y), &
+ fft_resolution(c_X),main_comm, local_resolution(c_Y), &
+ local_offset(c_Y), local_resolution(c_X),local_offset(c_X));
+
+ ! allocate local buffer (used to save datain/dataout1
+ ! ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
+ ! link datain and dataout1 to cbuffer, setting the right dimensions
+ call c_f_pointer(cbuffer1, datain1, &
+ [fft_resolution(c_X),local_resolution(c_Y)])
+@@ -115,17 +115,17 @@ contains
+ ! into dataout2 (input for backward transform and filter)
+ ! and to save (in-place) the transform of the second component
+ ! of the velocity
+- cbuffer2 = fftw_alloc_complex(alloc_local)
++ cbuffer2 = fftwf_alloc_complex(alloc_local)
+ call c_f_pointer(cbuffer2, datain2,&
+ [fft_resolution(c_X),local_resolution(c_Y)])
+ call c_f_pointer(cbuffer2, dataout2, [fft_resolution(c_Y),local_resolution(c_X)])
+
+ ! create MPI plan for in-place forward/backward DFT (note dimension reversal)
+- plan_forward1 = fftw_mpi_plan_dft_2d(fft_resolution(c_Y), fft_resolution(c_X),datain1,dataout1,&
++ plan_forward1 = fftwf_mpi_plan_dft_2d(fft_resolution(c_Y), fft_resolution(c_X),datain1,dataout1,&
+ main_comm,FFTW_FORWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_dft_2d(fft_resolution(c_Y),fft_resolution(c_X),dataout1,datain1,&
++ plan_backward1 = fftwf_mpi_plan_dft_2d(fft_resolution(c_Y),fft_resolution(c_X),dataout1,datain1,&
+ main_comm,FFTW_BACKWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_backward2 = fftw_mpi_plan_dft_2d(fft_resolution(c_Y),fft_resolution(c_X),dataout2,datain2,&
++ plan_backward2 = fftwf_mpi_plan_dft_2d(fft_resolution(c_Y),fft_resolution(c_X),dataout2,datain2,&
+ main_comm,FFTW_BACKWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+
+ call computeKxC(lengths(c_X))
+@@ -154,7 +154,7 @@ contains
+ end do
+
+ ! compute transform (as many times as desired)
+- call fftw_mpi_execute_dft(plan_forward1, datain1, dataout1)
++ call fftwf_mpi_execute_dft(plan_forward1, datain1, dataout1)
+
+ !!$ do i = 1, fft_resolution(c_Y)
+ !!$ write(*,'(a,i5,a,16f10.4)') 'out[',rank,'] ', dataout1(i,1:local_resolution(c_X))
+@@ -162,8 +162,8 @@ contains
+ !!$
+ call filter_poisson_2d()
+
+- call fftw_mpi_execute_dft(plan_backward1, dataout1, datain1)
+- call fftw_mpi_execute_dft(plan_backward2,dataout2,datain2)
++ call fftwf_mpi_execute_dft(plan_backward1, dataout1, datain1)
++ call fftwf_mpi_execute_dft(plan_backward2,dataout2,datain2)
+ do j = 1, local_resolution(c_Y)
+ do i = 1, fft_resolution(c_X)
+ velocity_x(i,j) = datain1(i,j)*normFFT
+@@ -198,7 +198,7 @@ contains
+ if(is2DUpToDate) return
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+
+@@ -206,11 +206,11 @@ contains
+ fft_resolution(:) = resolution(:)-1
+ halfLength = fft_resolution(c_X)/2+1
+ ! allocate local buffer (used to save datain/dataout1 ==> in-place transform!!)
+- alloc_local = fftw_mpi_local_size_2d_transposed(fft_resolution(c_Y),halfLength,main_comm,local_resolution(c_Y),&
++ alloc_local = fftwf_mpi_local_size_2d_transposed(fft_resolution(c_Y),halfLength,main_comm,local_resolution(c_Y),&
+ local_offset(c_Y),local_resolution(c_X),local_offset(c_X));
+
+ ! allocate local buffer (used to save datain/dataout1 ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
+
+ ! link rdatain1 and dataout1 to cbuffer, setting the right dimensions for each
+ call c_f_pointer(cbuffer1, rdatain1, [2*halfLength,local_resolution(c_Y)])
+@@ -218,17 +218,17 @@ contains
+
+ ! second buffer used for backward transform. Used to copy dataout1 into dataout2 (input for backward transform and filter)
+ ! and to save (in-place) the transform of the second component of the velocity
+- cbuffer2 = fftw_alloc_complex(alloc_local)
++ cbuffer2 = fftwf_alloc_complex(alloc_local)
+
+ call c_f_pointer(cbuffer2, rdatain2, [2*halfLength,local_resolution(c_Y)])
+ call c_f_pointer(cbuffer2, dataout2, [fft_resolution(c_Y),local_resolution(c_X)])
+
+ ! create MPI plans for in-place forward/backward DFT (note dimension reversal)
+- plan_forward1 = fftw_mpi_plan_dft_r2c_2d(fft_resolution(c_Y), fft_resolution(c_X), rdatain1, dataout1, &
++ plan_forward1 = fftwf_mpi_plan_dft_r2c_2d(fft_resolution(c_Y), fft_resolution(c_X), rdatain1, dataout1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
++ plan_backward1 = fftwf_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_backward2 = fftw_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
++ plan_backward2 = fftwf_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+
+ call computeKx(lengths(c_X))
+@@ -263,7 +263,7 @@ contains
+ !!$ end do
+ !!$
+ ! compute transform (as many times as desired)
+- call fftw_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
++ call fftwf_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
+
+ !!$ do i = 1, fft_resolution(c_Y)
+ !!$ write(*,'(a,i5,a,16f10.4)') 'aaaa[',rank,'] ', dataout1(i,1:local_resolution(c_X))
+@@ -276,8 +276,8 @@ contains
+ real(mk),dimension(:,:),intent(inout) :: velocity_x,velocity_y
+ integer(C_INTPTR_T) :: i, j
+
+- call fftw_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
+- call fftw_mpi_execute_dft_c2r(plan_backward2,dataout2,rdatain2)
++ call fftwf_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
++ call fftwf_mpi_execute_dft_c2r(plan_backward2,dataout2,rdatain2)
+ do j = 1, local_resolution(c_Y)
+ do i = 1, fft_resolution(c_X)
+ velocity_x(i,j) = rdatain1(i,j)*normFFT
+@@ -303,7 +303,7 @@ contains
+ real(mk),dimension(:,:),intent(inout) :: omega
+ integer(C_INTPTR_T) :: i, j
+
+- call fftw_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
++ call fftwf_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
+ do j = 1, local_resolution(c_Y)
+ do i = 1, fft_resolution(c_X)
+ omega(i,j) = rdatain1(i,j)*normFFT
+@@ -405,7 +405,7 @@ contains
+ subroutine filter_poisson_2d()
+
+ integer(C_INTPTR_T) :: i, j
+- complex(C_DOUBLE_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: coeff
+ if(local_offset(c_X)==0) then
+ if(local_offset(c_Y) == 0) then
+ dataout1(1,1) = 0.0
+@@ -452,9 +452,9 @@ contains
+
+ subroutine filter_diffusion_2d(nudt)
+
+- real(C_DOUBLE), intent(in) :: nudt
++ real(C_FLOAT), intent(in) :: nudt
+ integer(C_INTPTR_T) :: i, j
+- complex(C_DOUBLE_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: coeff
+
+ do i = 1,local_resolution(c_X)
+ do j = 1, fft_resolution(c_Y)
+@@ -467,20 +467,20 @@ contains
+
+ !> Clean fftw context (free memory, plans ...)
+ subroutine cleanFFTW_2d()
+- call fftw_destroy_plan(plan_forward1)
+- call fftw_destroy_plan(plan_backward1)
+- !call fftw_destroy_plan(plan_forward2)
+- !call fftw_destroy_plan(plan_backward2)
+- call fftw_free(cbuffer1)
+- call fftw_free(cbuffer2)
+- call fftw_mpi_cleanup()
++ call fftwf_destroy_plan(plan_forward1)
++ call fftwf_destroy_plan(plan_backward1)
++ !call fftwf_destroy_plan(plan_forward2)
++ !call fftwf_destroy_plan(plan_backward2)
++ call fftwf_free(cbuffer1)
++ call fftwf_free(cbuffer2)
++ call fftwf_mpi_cleanup()
+ deallocate(fft_resolution)
+ if(rank==0) close(21)
+ end subroutine cleanFFTW_2d
+
+ subroutine fft2d_diagnostics(nbelem)
+ integer(C_INTPTR_T), intent(in) :: nbelem
+- complex(C_DOUBLE_COMPLEX) :: memoryAllocated
++ complex(C_FLOAT_COMPLEX) :: memoryAllocated
+ memoryAllocated = real(nbelem*sizeof(memoryAllocated),mk)*1e-6
+ write(*,'(a,i5,a,i12,f10.2)') '[',rank,'] size of each buffer (elements / memory in MB):', &
+ nbelem, memoryAllocated
+@@ -514,10 +514,10 @@ contains
+ integer(C_INTPTR_T), dimension(2) :: n
+
+ !> Field (real values) for fftw input
+- real(C_DOUBLE), pointer :: rdatain1Many(:,:,:)
++ real(C_FLOAT), pointer :: rdatain1Many(:,:,:)
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+ howmany = 1
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+
+@@ -527,12 +527,12 @@ contains
+ n(1) = fft_resolution(2)
+ n(2) = halfLength
+ ! allocate local buffer (used to save datain/dataout1 ==> in-place transform!!)
+- alloc_local = fftw_mpi_local_size_many_transposed(2,n,howmany,FFTW_MPI_DEFAULT_BLOCK,&
++ alloc_local = fftwf_mpi_local_size_many_transposed(2,n,howmany,FFTW_MPI_DEFAULT_BLOCK,&
+ FFTW_MPI_DEFAULT_BLOCK,main_comm,local_resolution(c_Y),&
+ local_offset(c_Y),local_resolution(c_X),local_offset(c_X));
+
+ ! allocate local buffer (used to save datain/dataout1 ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
+
+ ! link rdatain1 and dataout1 to cbuffer, setting the right dimensions for each
+ call c_f_pointer(cbuffer1, rdatain1Many, [howmany,2*halfLength,local_resolution(c_Y)])
+@@ -540,17 +540,17 @@ contains
+
+ ! second buffer used for backward transform. Used to copy dataout1 into dataout2 (input for backward transform and filter)
+ ! and to save (in-place) the transform of the second component of the velocity
+- cbuffer2 = fftw_alloc_complex(alloc_local)
++ cbuffer2 = fftwf_alloc_complex(alloc_local)
+
+ call c_f_pointer(cbuffer2, rdatain1Many, [howmany,2*halfLength,local_resolution(c_Y)])
+ call c_f_pointer(cbuffer2, dataout2, [fft_resolution(c_Y),local_resolution(c_X)])
+
+ ! create MPI plans for in-place forward/backward DFT (note dimension reversal)
+- plan_forward1 = fftw_mpi_plan_dft_r2c_2d(fft_resolution(c_Y), fft_resolution(c_X), rdatain1Many, dataout1, &
++ plan_forward1 = fftwf_mpi_plan_dft_r2c_2d(fft_resolution(c_Y), fft_resolution(c_X), rdatain1Many, dataout1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
++ plan_backward1 = fftwf_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_backward2 = fftw_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
++ plan_backward2 = fftwf_mpi_plan_dft_c2r_2d(fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+
+ call computeKx(lengths(c_X))
+diff --git src/fftw/fft3d.f90 src/fftw/fft3d.f90
+index 0ad428e..dcefd1f 100755
+--- src/fftw/fft3d.f90
++++ src/fftw/fft3d.f90
+@@ -40,15 +40,15 @@ module fft3d
+ type(C_PTR) :: cbuffer3
+ !! Note Franck : check if local declarations of datain/out works and improve perfs.
+ !> Field (complex values) for fftw input
+- complex(C_DOUBLE_COMPLEX), pointer :: datain1(:,:,:)=>NULL(), datain2(:,:,:)=>NULL(), datain3(:,:,:)=>NULL()
++ complex(C_FLOAT_COMPLEX), pointer :: datain1(:,:,:)=>NULL(), datain2(:,:,:)=>NULL(), datain3(:,:,:)=>NULL()
+ !> Field (real values) for fftw input (these are only pointers to the cbuffers)
+- real(C_DOUBLE), pointer :: rdatain1(:,:,:)=>NULL() ,rdatain2(:,:,:)=>NULL() ,rdatain3(:,:,:)=>NULL()
++ real(C_FLOAT), pointer :: rdatain1(:,:,:)=>NULL() ,rdatain2(:,:,:)=>NULL() ,rdatain3(:,:,:)=>NULL()
+ !> Field (real values) for fftw input in the fftw-many case
+- real(C_DOUBLE), pointer :: rdatain_many(:,:,:,:)=>NULL()
++ real(C_FLOAT), pointer :: rdatain_many(:,:,:,:)=>NULL()
+ !> Field (complex values) for fftw (forward) output
+- complex(C_DOUBLE_COMPLEX), pointer :: dataout1(:,:,:)=>NULL() ,dataout2(:,:,:)=>NULL() ,dataout3(:,:,:)=>NULL()
++ complex(C_FLOAT_COMPLEX), pointer :: dataout1(:,:,:)=>NULL() ,dataout2(:,:,:)=>NULL() ,dataout3(:,:,:)=>NULL()
+ !> Field (complex values) for fftw (forward) output in the fftw-many case
+- complex(C_DOUBLE_COMPLEX), pointer :: dataout_many(:,:,:,:)=>NULL()
++ complex(C_FLOAT_COMPLEX), pointer :: dataout_many(:,:,:,:)=>NULL()
+ !> GLOBAL number of points in each direction on which fft is applied (--> corresponds to "real" resolution - 1)
+ integer(C_INTPTR_T),pointer :: fft_resolution(:)=>NULL()
+ !> LOCAL number of points for fft
+@@ -56,15 +56,15 @@ module fft3d
+ !> Offset in the direction of distribution
+ integer(c_INTPTR_T),dimension(3) :: local_offset
+ !> wave numbers for fft in x direction
+- real(C_DOUBLE), pointer :: kx(:)
++ real(C_FLOAT), pointer :: kx(:)
+ !> wave numbers for fft in y direction
+- real(C_DOUBLE), pointer :: ky(:)
++ real(C_FLOAT), pointer :: ky(:)
+ !> wave numbers for fft in z direction
+- real(C_DOUBLE), pointer :: kz(:)
++ real(C_FLOAT), pointer :: kz(:)
+ !> log file for fftw
+ character(len=20),parameter :: filename ="parmesfftw.log"
+ !> normalization factor
+- real(C_DOUBLE) :: normFFT
++ real(C_FLOAT) :: normFFT
+ !> true if we use fftw-many routines
+ logical :: manycase
+ !> true if all the allocation stuff for global variables has been done.
+@@ -89,7 +89,7 @@ contains
+ if(is3DUpToDate) return
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+
+@@ -98,7 +98,7 @@ contains
+ fft_resolution(:) = resolution(:)-1
+
+ ! compute "optimal" size (according to fftw) for local data (warning : dimension reversal)
+- alloc_local = fftw_mpi_local_size_3d_transposed(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),main_comm,&
++ alloc_local = fftwf_mpi_local_size_3d_transposed(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),main_comm,&
+ local_resolution(c_Z),local_offset(c_Z),local_resolution(c_Y),local_offset(c_Y));
+
+ ! Set a default value for c_X components.
+@@ -106,35 +106,35 @@ contains
+ local_resolution(c_X) = fft_resolution(c_X)
+
+ ! allocate local buffer (used to save datain/dataout ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
+ ! link datain and dataout to cbuffer, setting the right dimensions for each
+ call c_f_pointer(cbuffer1, datain1, [fft_resolution(c_X),fft_resolution(c_Y),local_resolution(c_Z)])
+ call c_f_pointer(cbuffer1, dataout1, [fft_resolution(c_X),fft_resolution(c_Z),local_resolution(c_Y)])
+
+ ! second buffer used for backward transform. Used to copy dataout into dataout2 (input for backward transform and filter)
+ ! and to save (in-place) the transform of the second component of the velocity
+- cbuffer2 = fftw_alloc_complex(alloc_local)
++ cbuffer2 = fftwf_alloc_complex(alloc_local)
+ call c_f_pointer(cbuffer2, datain2, [fft_resolution(c_X),fft_resolution(c_Y),local_resolution(c_Z)])
+ call c_f_pointer(cbuffer2, dataout2, [fft_resolution(c_X),fft_resolution(c_Z),local_resolution(c_Y)])
+
+ ! second buffer used for backward transform. Used to copy dataout into dataout2 (input for backward transform and filter)
+ ! and to save (in-place) the transform of the second component of the velocity
+- cbuffer3 = fftw_alloc_complex(alloc_local)
++ cbuffer3 = fftwf_alloc_complex(alloc_local)
+ call c_f_pointer(cbuffer3, datain3, [fft_resolution(c_X),fft_resolution(c_Y),local_resolution(c_Z)])
+ call c_f_pointer(cbuffer3, dataout3, [fft_resolution(c_X),fft_resolution(c_Z),local_resolution(c_Y)])
+
+ ! create MPI plan for in-place forward/backward DFT (note dimension reversal)
+- plan_forward1 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain1,dataout1,&
++ plan_forward1 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain1,dataout1,&
+ main_comm,FFTW_FORWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout1,datain1,&
++ plan_backward1 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout1,datain1,&
+ main_comm,FFTW_BACKWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_forward2 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain2,dataout2,&
++ plan_forward2 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain2,dataout2,&
+ main_comm,FFTW_FORWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward2 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout2,datain2,&
++ plan_backward2 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout2,datain2,&
+ main_comm,FFTW_BACKWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_forward3 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain3,dataout3,&
++ plan_forward3 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z), fft_resolution(c_Y), fft_resolution(c_X),datain3,dataout3,&
+ main_comm,FFTW_FORWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward3 = fftw_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout3,datain3,&
++ plan_backward3 = fftwf_mpi_plan_dft_3d(fft_resolution(c_Z),fft_resolution(c_Y),fft_resolution(c_X),dataout3,datain3,&
+ main_comm,FFTW_BACKWARD,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+
+ call computeKx(lengths(c_X))
+@@ -173,17 +173,17 @@ contains
+ end do
+ end do
+ ! compute transform (as many times as desired)
+- call fftw_mpi_execute_dft(plan_forward1, datain1, dataout1)
+- call fftw_mpi_execute_dft(plan_forward2, datain2, dataout2)
+- call fftw_mpi_execute_dft(plan_forward3, datain3, dataout3)
++ call fftwf_mpi_execute_dft(plan_forward1, datain1, dataout1)
++ call fftwf_mpi_execute_dft(plan_forward2, datain2, dataout2)
++ call fftwf_mpi_execute_dft(plan_forward3, datain3, dataout3)
+
+ ! apply poisson filter
+ call filter_poisson_3d()
+
+ ! inverse transform to retrieve velocity
+- call fftw_mpi_execute_dft(plan_backward1, dataout1,datain1)
+- call fftw_mpi_execute_dft(plan_backward2,dataout2,datain2)
+- call fftw_mpi_execute_dft(plan_backward3,dataout3,datain3)
++ call fftwf_mpi_execute_dft(plan_backward1, dataout1,datain1)
++ call fftwf_mpi_execute_dft(plan_backward2,dataout2,datain2)
++ call fftwf_mpi_execute_dft(plan_backward3,dataout3,datain3)
+ do k =1, local_resolution(c_Z)
+ do j = 1, fft_resolution(c_Y)
+ do i = 1, fft_resolution(c_X)
+@@ -213,7 +213,7 @@ contains
+ if(is3DUpToDate) return
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+ allocate(fft_resolution(3))
+@@ -221,7 +221,7 @@ contains
+ halfLength = fft_resolution(c_X)/2+1
+
+ ! compute "optimal" size (according to fftw) for local data (warning : dimension reversal)
+- alloc_local = fftw_mpi_local_size_3d_transposed(fft_resolution(c_Z),fft_resolution(c_Y),halfLength,&
++ alloc_local = fftwf_mpi_local_size_3d_transposed(fft_resolution(c_Z),fft_resolution(c_Y),halfLength,&
+ main_comm,local_resolution(c_Z),local_offset(c_Z),local_resolution(c_Y),local_offset(c_Y));
+
+ ! init c_X part. This is required to compute kx with the same function in 2d and 3d cases.
+@@ -229,9 +229,9 @@ contains
+ local_resolution(c_X) = halfLength
+
+ ! allocate local buffer (used to save datain/dataout ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
+- cbuffer2 = fftw_alloc_complex(alloc_local)
+- cbuffer3 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
++ cbuffer2 = fftwf_alloc_complex(alloc_local)
++ cbuffer3 = fftwf_alloc_complex(alloc_local)
+
+ ! link rdatain and dataout to cbuffer, setting the right dimensions for each
+ call c_f_pointer(cbuffer1, rdatain1, [2*halfLength,fft_resolution(c_Y),local_resolution(c_Z)])
+@@ -246,17 +246,17 @@ contains
+ rdatain3 = 0.0
+
+ ! create MPI plans for in-place forward/backward DFT (note dimension reversal)
+- plan_forward1 = fftw_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain1, dataout1, &
++ plan_forward1 = fftwf_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain1, dataout1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
++ plan_backward1 = fftwf_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout1, rdatain1, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_forward2 = fftw_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain2, dataout2, &
++ plan_forward2 = fftwf_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain2, dataout2, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward2 = fftw_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
++ plan_backward2 = fftwf_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout2, rdatain2, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+- plan_forward3 = fftw_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain3, dataout3, &
++ plan_forward3 = fftwf_mpi_plan_dft_r2c_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), rdatain3, dataout3, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward3 = fftw_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout3, rdatain3, &
++ plan_backward3 = fftwf_mpi_plan_dft_c2r_3d(fft_resolution(c_Z),fft_resolution(c_Y), fft_resolution(c_X), dataout3, rdatain3, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+
+ call computeKx(lengths(c_X))
+@@ -279,7 +279,7 @@ contains
+
+ real(mk),dimension(:,:,:),intent(in) :: omega_x,omega_y,omega_z
+ integer, dimension(3), intent(in) :: ghosts
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k, ig, jg, kg
+
+ ! ig, jg, kg are used to take into account
+@@ -301,9 +301,9 @@ contains
+
+ ! compute transforms for each component
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
+- call fftw_mpi_execute_dft_r2c(plan_forward2, rdatain2, dataout2)
+- call fftw_mpi_execute_dft_r2c(plan_forward3, rdatain3, dataout3)
++ call fftwf_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
++ call fftwf_mpi_execute_dft_r2c(plan_forward2, rdatain2, dataout2)
++ call fftwf_mpi_execute_dft_r2c(plan_forward3, rdatain3, dataout3)
+ !!print *, "r2c time = ", MPI_WTIME() - start
+
+ end subroutine r2c_3d
+@@ -316,12 +316,12 @@ contains
+ subroutine c2r_3d(velocity_x,velocity_y,velocity_z, ghosts)
+ real(mk),dimension(:,:,:),intent(inout) :: velocity_x,velocity_y,velocity_z
+ integer, dimension(3), intent(in) :: ghosts
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k, ig, jg, kg
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
+- call fftw_mpi_execute_dft_c2r(plan_backward2,dataout2,rdatain2)
+- call fftw_mpi_execute_dft_c2r(plan_backward3,dataout3,rdatain3)
++ call fftwf_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
++ call fftwf_mpi_execute_dft_c2r(plan_backward2,dataout2,rdatain2)
++ call fftwf_mpi_execute_dft_c2r(plan_backward3,dataout3,rdatain3)
+ !! print *, "c2r time : ", MPI_WTIME() -start
+ ! copy back to velocity and normalisation
+ do k =1, local_resolution(c_Z)
+@@ -346,7 +346,7 @@ contains
+
+ real(mk),dimension(:,:,:),intent(in) :: omega
+ integer, dimension(3), intent(in) :: ghosts
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k, ig, jg, kg
+
+ ! ig, jg, kg are used to take into account
+@@ -366,7 +366,7 @@ contains
+
+ ! compute transforms for each component
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
++ call fftwf_mpi_execute_dft_r2c(plan_forward1, rdatain1, dataout1)
+ !!print *, "r2c time = ", MPI_WTIME() - start
+
+ end subroutine r2c_scalar_3d
+@@ -377,10 +377,10 @@ contains
+ subroutine c2r_scalar_3d(omega, ghosts)
+ real(mk),dimension(:,:,:),intent(inout) :: omega
+ integer, dimension(3), intent(in) :: ghosts
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k, ig, jg, kg
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
++ call fftwf_mpi_execute_dft_c2r(plan_backward1,dataout1,rdatain1)
+ !! print *, "c2r time : ", MPI_WTIME() -start
+ ! copy back to velocity and normalisation
+ do k =1, local_resolution(c_Z)
+@@ -412,7 +412,7 @@ contains
+ integer(C_INTPTR_T),dimension(3) :: n
+
+ ! init fftw mpi context
+- call fftw_mpi_init()
++ call fftwf_mpi_init()
+ blocksize = FFTW_MPI_DEFAULT_BLOCK
+ if(rank==0) open(unit=21,file=filename,form="formatted")
+ allocate(fft_resolution(3))
+@@ -423,7 +423,7 @@ contains
+ n(3) = halfLength
+ howmany = 3
+ ! compute "optimal" size (according to fftw) for local data (warning : dimension reversal)
+- alloc_local = fftw_mpi_local_size_many_transposed(3,n,howmany,blocksize,blocksize,&
++ alloc_local = fftwf_mpi_local_size_many_transposed(3,n,howmany,blocksize,blocksize,&
+ main_comm,local_resolution(c_Z),local_offset(c_Z),local_resolution(c_Y),local_offset(c_Y));
+
+ ! init c_X part. This is required to compute kx with the same function in 2d and 3d cases.
+@@ -431,7 +431,7 @@ contains
+ local_resolution(c_X) = halfLength
+
+ ! allocate local buffer (used to save datain/dataout ==> in-place transform!!)
+- cbuffer1 = fftw_alloc_complex(alloc_local)
++ cbuffer1 = fftwf_alloc_complex(alloc_local)
+
+ ! link rdatain and dataout to cbuffer, setting the right dimensions for each
+ call c_f_pointer(cbuffer1, rdatain_many, [howmany,2*halfLength,fft_resolution(c_Y),local_resolution(c_Z)])
+@@ -440,9 +440,9 @@ contains
+ ! create MPI plans for in-place forward/backward DFT (note dimension reversal)
+ n(3) = fft_resolution(c_X)
+
+- plan_forward1 = fftw_mpi_plan_many_dft_r2c(3,n,howmany,blocksize,blocksize, rdatain_many, dataout_many, &
++ plan_forward1 = fftwf_mpi_plan_many_dft_r2c(3,n,howmany,blocksize,blocksize, rdatain_many, dataout_many, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_OUT))
+- plan_backward1 = fftw_mpi_plan_many_dft_c2r(3,n,howmany,blocksize,blocksize, dataout_many, rdatain_many, &
++ plan_backward1 = fftwf_mpi_plan_many_dft_c2r(3,n,howmany,blocksize,blocksize, dataout_many, rdatain_many, &
+ main_comm,ior(FFTW_MEASURE,FFTW_MPI_TRANSPOSED_IN))
+ call computeKx(lengths(c_X))
+ call computeKy(lengths(c_Y))
+@@ -464,7 +464,7 @@ contains
+ subroutine r2c_3d_many(omega_x,omega_y,omega_z)
+
+ real(mk),dimension(:,:,:),intent(in) :: omega_x,omega_y,omega_z
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k
+
+ ! init
+@@ -480,7 +480,7 @@ contains
+
+ ! compute transform (as many times as desired)
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_r2c(plan_forward1, rdatain_many, dataout_many)
++ call fftwf_mpi_execute_dft_r2c(plan_forward1, rdatain_many, dataout_many)
+ !! print *, "r2c time = ", MPI_WTIME() - start
+
+ end subroutine r2c_3d_many
+@@ -491,11 +491,11 @@ contains
+ !! @param[in,out] velocity_z 3d scalar field, z-component of the output vector field
+ subroutine c2r_3d_many(velocity_x,velocity_y,velocity_z)
+ real(mk),dimension(:,:,:),intent(inout) :: velocity_x,velocity_y,velocity_z
+- real(mk) :: start
++ real(8) :: start
+ integer(C_INTPTR_T) :: i,j,k
+
+ start = MPI_WTIME()
+- call fftw_mpi_execute_dft_c2r(plan_backward1,dataout_many,rdatain_many)
++ call fftwf_mpi_execute_dft_c2r(plan_backward1,dataout_many,rdatain_many)
+ !! print *, "c2r time : ", MPI_WTIME() -start
+ do k =1, local_resolution(c_Z)
+ do j = 1, fft_resolution(c_Y)
+@@ -536,7 +536,7 @@ contains
+ !> Computation of frequencies coeff, over distributed direction(s)
+ !> @param lengths size of the domain
+ subroutine computeKy(length)
+- real(C_DOUBLE), intent(in) :: length
++ real(C_FLOAT), intent(in) :: length
+
+ !! Local loops indices
+ integer(C_INTPTR_T) :: i
+@@ -589,8 +589,8 @@ contains
+ subroutine filter_poisson_3d()
+
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2
+
+ ! Set first coeff (check for "all freq = 0" case)
+ if(local_offset(c_Y) == 0) then
+@@ -648,10 +648,10 @@ contains
+ !! @param[in] nudt \f$ \nu\times dt\f$, diffusion coefficient times current time step
+ subroutine filter_curl_diffusion_3d(nudt)
+
+- real(C_DOUBLE), intent(in) :: nudt
++ real(C_FLOAT), intent(in) :: nudt
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2
+
+ !! mind the transpose -> index inversion between y and z
+ do j = 1,local_resolution(c_Y)
+@@ -674,9 +674,9 @@ contains
+ !! @param[in] nudt \f$ \nu\times dt\f$, diffusion coefficient times current time step
+ subroutine filter_diffusion_3d(nudt)
+
+- real(C_DOUBLE), intent(in) :: nudt
++ real(C_FLOAT), intent(in) :: nudt
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: coeff
+
+ !! mind the transpose -> index inversion between y and z
+ do j = 1,local_resolution(c_Y)
+@@ -697,8 +697,8 @@ contains
+ subroutine filter_curl_3d()
+
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2
+
+ !! mind the transpose -> index inversion between y and z
+ do j = 1,local_resolution(c_Y)
+@@ -721,8 +721,8 @@ contains
+ subroutine filter_projection_om_3d()
+
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2,buffer3
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2,buffer3
+
+ ! Set first coeff (check for "all freq = 0" case)
+ if(local_offset(c_Y) == 0) then
+@@ -783,9 +783,9 @@ contains
+ !! @param[in] dxf, dyf, dzf: grid filter size = domainLength/(CoarseRes-1)
+ subroutine filter_multires_om_3d(dxf, dyf, dzf)
+
+- real(C_DOUBLE), intent(in) :: dxf, dyf, dzf
++ real(C_FLOAT), intent(in) :: dxf, dyf, dzf
+ integer(C_INTPTR_T) :: i,j,k
+- real(C_DOUBLE) :: kxc, kyc, kzc
++ real(C_FLOAT) :: kxc, kyc, kzc
+
+ kxc = pi / dxf
+ kyc = pi / dyf
+@@ -810,7 +810,7 @@ contains
+ !! pressure from velocity in the Fourier space
+ subroutine filter_pressure_3d()
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: coeff
+
+ ! Set first coeff (check for "all freq = 0" case)
+ if(local_offset(c_Y) == 0) then
+@@ -851,8 +851,8 @@ contains
+ subroutine filter_poisson_3d_many()
+
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2
+
+ ! Set first coeff (check for "all freq = 0" case)
+ if(local_offset(c_Y) == 0) then
+@@ -908,10 +908,10 @@ contains
+ !! @param[in] nudt \f$ \nu\times dt\f$, diffusion coefficient times current time step
+ subroutine filter_diffusion_3d_many(nudt)
+
+- real(C_DOUBLE), intent(in) :: nudt
++ real(C_FLOAT), intent(in) :: nudt
+ integer(C_INTPTR_T) :: i,j,k
+- complex(C_DOUBLE_COMPLEX) :: coeff
+- complex(C_DOUBLE_COMPLEX) :: buffer1,buffer2
++ complex(C_FLOAT_COMPLEX) :: coeff
++ complex(C_FLOAT_COMPLEX) :: buffer1,buffer2
+
+ !! mind the transpose -> index inversion between y and z
+ do j = 1,local_resolution(c_Y)
+@@ -931,18 +931,18 @@ contains
+
+ !> Clean fftw context (free memory, plans ...)
+ subroutine cleanFFTW_3d()
+- call fftw_destroy_plan(plan_forward1)
+- call fftw_destroy_plan(plan_backward1)
++ call fftwf_destroy_plan(plan_forward1)
++ call fftwf_destroy_plan(plan_backward1)
+ if(.not.manycase) then
+- call fftw_destroy_plan(plan_forward2)
+- call fftw_destroy_plan(plan_backward2)
+- call fftw_destroy_plan(plan_forward3)
+- call fftw_destroy_plan(plan_backward3)
+- call fftw_free(cbuffer2)
+- call fftw_free(cbuffer3)
++ call fftwf_destroy_plan(plan_forward2)
++ call fftwf_destroy_plan(plan_backward2)
++ call fftwf_destroy_plan(plan_forward3)
++ call fftwf_destroy_plan(plan_backward3)
++ call fftwf_free(cbuffer2)
++ call fftwf_free(cbuffer3)
+ endif
+- call fftw_free(cbuffer1)
+- call fftw_mpi_cleanup()
++ call fftwf_free(cbuffer1)
++ call fftwf_mpi_cleanup()
+ deallocate(fft_resolution)
+ deallocate(kx,ky,kz)
+ if(rank==0) close(21)
+@@ -953,7 +953,7 @@ contains
+ integer(C_INTPTR_T), intent(in) :: nbelem
+ ! number of buffers used for fftw
+ integer, optional,intent(in) :: howmany
+- complex(C_DOUBLE_COMPLEX) :: memoryAllocated
++ complex(C_FLOAT_COMPLEX) :: memoryAllocated
+
+ integer :: nbFields
+ if(present(howmany)) then
+diff --git src/main/main.f90 src/main/main.f90
+index 38cc6a5..0fc8f5f 100755
+--- src/main/main.f90
++++ src/main/main.f90
+@@ -9,7 +9,7 @@ use vectorcalculus
+ implicit none
+
+ integer :: info
+-real(mk) :: start, end
++real(8) :: start, end
+
+ !complex(mk), dimension(resolution(1),resolution(2)) :: omega,velocity_x,velocity_y
+ call MPI_Init(info)
+@@ -116,7 +116,8 @@ contains
+ real(mk),dimension(3) :: lengths,step
+ integer, dimension(3) :: ghosts_v, ghosts_w
+ integer(C_INTPTR_T),dimension(3) :: nfft,offset
+- real(mk) :: error,start
++ real(mk) :: error
++ real(8) :: start
+ logical :: ok
+
+ if (rank==0) print *, " ======= Test 3D Poisson (r2c) solver for resolution ", resolution
+diff --git src/scalesInterface/precision_tools.f90 src/scalesInterface/precision_tools.f90
+index 5dacf3f..4ae1a6a 100644
+--- src/scalesInterface/precision_tools.f90
++++ src/scalesInterface/precision_tools.f90
+@@ -20,15 +20,15 @@
+ !------------------------------------------------------------------------------
+
+ MODULE precision_tools
+- use mpi, only: MPI_DOUBLE_PRECISION
++ use mpi, only: MPI_REAL
+ implicit None
+
+ !> Floats precision
+ INTEGER, PARAMETER :: SP = kind(1.0)
+ INTEGER, PARAMETER :: DP = kind(1.0d0)
+- INTEGER, PARAMETER :: WP = DP
++ INTEGER, PARAMETER :: WP = SP
+ !> the MPI type for REAL exchanges in simple or double precision
+- INTEGER, parameter :: MPI_REAL_WP = MPI_DOUBLE_PRECISION
++ INTEGER, parameter :: MPI_REAL_WP = MPI_REAL
+ REAL(WP), PRIVATE :: sample_real_at_WP
+ REAL(WP), PARAMETER :: MAX_REAL_WP = HUGE(sample_real_at_WP)
+ INTEGER, PRIVATE :: sample_int
--
GitLab