diff --git a/mex/build/mjdgges.am b/mex/build/mjdgges.am
index 6a238c9c6..d9e05a2d2 100644
--- a/mex/build/mjdgges.am
+++ b/mex/build/mjdgges.am
@@ -1,9 +1,11 @@
 mex_PROGRAMS = mjdgges
 
-nodist_mjdgges_SOURCES = mjdgges.cc
+nodist_mjdgges_SOURCES = mjdgges.F08 matlab_mex.F08 blas_lapack.F08
 
 BUILT_SOURCES = $(nodist_mjdgges_SOURCES)
 CLEANFILES = $(nodist_mjdgges_SOURCES)
 
-%.cc: $(top_srcdir)/../../sources/mjdgges/%.cc
+mjdgges.o : matlab_mex.mod lapack.mod
+
+%.F08: $(top_srcdir)/../../sources/mjdgges/%.F08
 	$(LN_S) -f $< $@
diff --git a/mex/sources/blas_lapack.F08 b/mex/sources/blas_lapack.F08
index 440e10513..a93aff118 100644
--- a/mex/sources/blas_lapack.F08
+++ b/mex/sources/blas_lapack.F08
@@ -1,4 +1,4 @@
-! Copyright © 2019 Dynare Team
+! Copyright © 2019-2020 Dynare Team
 !
 ! This file is part of Dynare.
 !
@@ -20,8 +20,10 @@ module blas
 
 #if defined(MATLAB_MEX_FILE) && __SIZEOF_POINTER__ == 8
   integer, parameter :: blint = int64
+  integer, parameter :: bllog = 8 ! Logical kind, gfortran-specific
 #else
   integer, parameter :: blint = int32
+  integer, parameter :: bllog = 4 ! Logical kind, gfortran-specific
 #endif
 
   interface
@@ -43,10 +45,29 @@ module lapack
      subroutine dgesv(n, nrhs, a, lda, ipiv, b, ldb, info)
        import :: blint, real64
        integer(blint), intent(in) :: n, nrhs, lda, ldb
-       real(real64), dimension(*), intent(inout) :: a
-       real(real64), dimension(*), intent(inout) :: b
+       real(real64), dimension(*), intent(inout) :: a, b
        integer(blint), dimension(*), intent(out) :: ipiv
        integer(blint), intent(out) :: info
      end subroutine dgesv
   end interface
+
+  interface
+     subroutine dgges(jobvsl, jobvsr, sort, selctg, n, a, lda, b, ldb, sdim, &
+          alphar, alphai, beta, vsl, ldvsl, vsr, ldvsr, work, lwork, bwork, &
+          info)
+       import :: blint, bllog, real64
+       character :: jobvsl, jobvsr, sort
+       interface
+          logical(bllog) function selctg(alphar, alphai, beta)
+            import :: bllog, real64
+            real(real64), intent(in) :: alphar, alphai, beta
+          end function selctg
+       end interface
+       integer(blint), intent(in) :: n, lda, ldb, ldvsl, ldvsr, lwork
+       real(real64), dimension(*), intent(inout) :: a, b
+       real(real64), dimension(*), intent(out) :: alphar, alphai, beta, vsl, vsr, work
+       logical(bllog), dimension(*), intent(out) :: bwork
+       integer(blint), intent(out) :: sdim, info
+     end subroutine dgges
+  end interface
 end module lapack
diff --git a/mex/sources/defines.F08 b/mex/sources/defines.F08
new file mode 100644
index 000000000..80d607f63
--- /dev/null
+++ b/mex/sources/defines.F08
@@ -0,0 +1 @@
+#define MX_HAS_INTERLEAVED_COMPLEX (defined(MATLAB_MEX_FILE) && MATLAB_VERSION >= 0x0904)
diff --git a/mex/sources/matlab_mex.F08 b/mex/sources/matlab_mex.F08
index 1fbbf8ba9..7a853f710 100644
--- a/mex/sources/matlab_mex.F08
+++ b/mex/sources/matlab_mex.F08
@@ -52,7 +52,7 @@
 # define API_VER2 ""
 #endif
 
-#define MX_HAS_INTERLEAVED_COMPLEX (defined(MATLAB_MEX_FILE) && MATLAB_VERSION >= 0x0904)
+#include "defines.F08"
 
 !!! C Matrix API
 !!! Listed in same order as https://fr.mathworks.com/help/matlab/cc-mx-matrix-library.html
diff --git a/mex/sources/mjdgges/mjdgges.F08 b/mex/sources/mjdgges/mjdgges.F08
new file mode 100644
index 000000000..dbc5da9d1
--- /dev/null
+++ b/mex/sources/mjdgges/mjdgges.F08
@@ -0,0 +1,167 @@
+! Copyright © 2006-2020 Dynare Team
+!
+! This file is part of Dynare.
+!
+! Dynare 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.
+!
+! Dynare 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 Dynare.  If not, see <http://www.gnu.org/licenses/>.
+
+#include "defines.F08"
+
+module select_fct_mod
+  use iso_fortran_env
+  implicit none
+
+  real(real64) :: criterium
+contains
+  logical(bllog) function select_fct(alpha_r, alpha_i, beta)
+    use blas
+
+    real(real64), intent(in) :: alpha_r, alpha_i, beta
+
+    select_fct = alpha_r**2 + alpha_i**2 < criterium**2 * beta**2
+  end function select_fct
+end module select_fct_mod
+
+subroutine mexFunction(nlhs, plhs, nrhs, prhs) bind(c, name='mexFunction')
+  use iso_fortran_env
+  use iso_c_binding
+  use select_fct_mod
+  use matlab_mex
+  use lapack
+  implicit none
+
+  type(c_ptr), dimension(*), intent(in), target :: prhs
+  type(c_ptr), dimension(*), intent(out) :: plhs
+  integer(c_int), intent(in), value :: nlhs, nrhs
+
+  integer(c_size_t) :: m1, n1, m2, n2
+  real(real64) :: zhreshold
+  integer(blint) :: n_bl, lwork, info_bl, sdim_bl
+  real(real64), dimension(:), allocatable :: alpha_r, alpha_i, beta, work
+  logical(bllog), dimension(:), allocatable :: bwork
+  real(real64), dimension(:), pointer :: s, t, z, info, sdim, vsl
+#if MX_HAS_INTERLEAVED_COMPLEX
+  complex(real64), dimension(:), pointer :: gev
+#else
+  real(real64), dimension(:), pointer :: gev_r, gev_i
+#endif
+#if defined(MATLAB_MEX_FILE) && MATLAB_VERSION < 0x0904
+  ! Workaround for MKL bug, see below
+  real(real64), dimension(:), allocatable, target :: vsl_target
+#endif
+
+  if (nrhs < 2 .or. nrhs > 4 .or. nlhs /= 7) then
+     call mexErrMsgTxt("MJDGGES: takes 2, 3 or 4 input arguments and exactly 7 output arguments.")
+     return
+  end if
+
+  m1 = mxGetM(prhs(1))
+  n1 = mxGetN(prhs(1))
+  m2 = mxGetM(prhs(2))
+  n2 = mxGetN(prhs(2))
+
+  if (.not. mxIsDouble(prhs(1)) .or. mxIsComplex(prhs(1)) &
+      .or. .not. mxIsDouble(prhs(2)) .or. mxIsComplex(prhs(2)) &
+      .or. m1 /= n1 .or. m2 /= n1 .or. m2 /= n2) then
+     call mexErrMsgTxt("MJDGGES requires two square real matrices of the same dimension.")
+     return
+  end if
+
+  ! Set criterium for stable eigenvalues
+  if (nrhs >= 3 .and. mxGetM(prhs(3)) > 0) then
+     associate (crit_arg => mxGetPr(prhs(3)))
+       criterium = crit_arg(1)
+     end associate
+  else
+     criterium = 1_real64 + 1e-6_real64
+  end if
+
+  ! set criterium for 0/0 generalized eigenvalues */
+  if (nrhs == 4 .and. mxGetM(prhs(4)) > 0) then
+     associate (zhresh_arg => mxGetPr(prhs(4)))
+       zhreshold = zhresh_arg(1)
+     end associate
+  else
+     zhreshold = 1e-6_real64
+  end if
+
+  plhs(2) = mxCreateDoubleMatrix(n1, n1, mxREAL)
+  plhs(3) = mxCreateDoubleMatrix(n1, n1, mxREAL)
+  plhs(4) = mxCreateDoubleMatrix(n1, n1, mxREAL)
+  plhs(5) = mxCreateDoubleMatrix(1_mwSize, 1_mwSize, mxREAL)
+  plhs(6) = mxCreateDoubleMatrix(n1, 1_mwSize, mxCOMPLEX)
+  plhs(7) = mxCreateDoubleMatrix(1_mwSize, 1_mwSize, mxREAL)
+
+  s => mxGetPr(plhs(2))
+  t => mxGetPr(plhs(3))
+  sdim => mxGetPr(plhs(5))
+#if MX_HAS_INTERLEAVED_COMPLEX
+  gev => mxGetComplexDoubles(plhs(6))
+#else
+  gev_r => mxGetPr(plhs(6))
+  gev_i => mxGetPi(plhs(6))
+#endif
+  info => mxGetPr(plhs(7))
+  z => mxGetPr(plhs(4))
+#if defined(MATLAB_MEX_FILE) && MATLAB_VERSION < 0x0904
+  ! The left Schur vectors (VSL) are normally not computed, since JOBVSL="N".
+  ! But old MKL versions (at least the one shipped with MATLAB R2009b/7.9,
+  ! which is MKL 10.1) are buggy, and passing nullptr for VSL leads to a crash.
+  ! Hence we need to allocate space for it.
+  ! The bug seems to be fixed in MATLAB R2010a/7.10 (MKL 10.2), but we use the
+  ! workaround for all versions < R2018a/9.4, since those share the same
+  ! ABI and hence the same executables.
+  allocate(vsl_target(n1*n1))
+  vsl => vsl_target
+#else
+  vsl => null()
+#endif
+
+  ! Copy input matrices, since we can’t modify them
+  associate (a => mxGetPr(prhs(1)), b => mxGetPr(prhs(2)))
+    s = a
+    t = b
+  end associate
+
+  n_bl = int(n1, blint)
+  lwork = 16*n_bl + 16
+  allocate(alpha_r(n_bl), alpha_i(n_bl), beta(n_bl), bwork(n_bl), work(lwork))
+
+  call dgges("N", "V", "S", select_fct, n_bl, s, n_bl, t, n_bl, sdim_bl, &
+       alpha_r, alpha_i, beta, vsl, n_bl, z, n_bl, work, lwork, bwork, info_bl)
+
+  info = info_bl
+  sdim = sdim_bl
+
+#if MX_HAS_INTERLEAVED_COMPLEX
+  where (alpha_i == 0_real64 .and. beta == 0_real64)
+     gev = alpha_r / beta
+  elsewhere
+     gev = cmplx(alpha_r, alpha_i, real64) / beta
+  end where
+#else
+  gev_r = alpha_r / beta
+  where (alpha_i == 0_real64 .and. beta == 0_real64)
+     gev_i = 0_real64
+  elsewhere
+     gev_i = alpha_i / beta
+  end where
+#endif
+
+  ! If the ratio of some eigenvalue is too close to 0/0, return specific
+  ! error number (only if no other error)
+  if (any(abs(alpha_r) <= zhreshold .and. abs(beta) <= zhreshold) .and. info_bl == 0) &
+       info = 30
+
+  plhs(1) = mxCreateDoubleScalar(0._real64)
+end subroutine mexFunction
diff --git a/mex/sources/mjdgges/mjdgges.cc b/mex/sources/mjdgges/mjdgges.cc
deleted file mode 100644
index b3873a90a..000000000
--- a/mex/sources/mjdgges/mjdgges.cc
+++ /dev/null
@@ -1,157 +0,0 @@
-/*
- * Copyright © 2006-2019 Dynare Team
- *
- * This file is part of Dynare.
- *
- * Dynare 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.
- *
- * Dynare 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 Dynare.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include <cmath>
-#include <algorithm>
-#include <memory>
-
-#include <dynmex.h>
-#include <dynlapack.h>
-
-double criterium;
-
-lapack_int
-my_criteria(const double *alpha_r, const double *alpha_i, const double *beta)
-{
-  return *alpha_r * *alpha_r + *alpha_i * *alpha_i < criterium * criterium * *beta * *beta;
-}
-
-/* MATLAB interface */
-void
-mexFunction(int nlhs, mxArray *plhs[],
-            int nrhs, const mxArray *prhs[])
-{
-  /* Check for proper number of arguments */
-  if (nrhs < 2 || nrhs > 4 || nlhs == 0 || nlhs > 7)
-    DYN_MEX_FUNC_ERR_MSG_TXT("MJDGGES: takes 2, 3 or 4 input arguments and between 1 and 7 output arguments.");
-
-  /* Check that A and B are real matrices of the same dimension.*/
-  size_t m1 = mxGetM(prhs[0]);
-  size_t n1 = mxGetN(prhs[0]);
-  size_t m2 = mxGetM(prhs[1]);
-  size_t n2 = mxGetN(prhs[1]);
-  if (!mxIsDouble(prhs[0]) || mxIsComplex(prhs[0])
-      || !mxIsDouble(prhs[1]) || mxIsComplex(prhs[1])
-      || m1 != n1 || m2 != n1 || m2 != n2)
-    DYN_MEX_FUNC_ERR_MSG_TXT("MJDGGES requires two square real matrices of the same dimension.");
-
-  /* Create a matrix for the return argument */
-  plhs[1] = mxCreateDoubleMatrix(n1, n1, mxREAL);
-  plhs[2] = mxCreateDoubleMatrix(n1, n1, mxREAL);
-  plhs[3] = mxCreateDoubleMatrix(n1, n1, mxREAL);
-  plhs[4] = mxCreateDoubleMatrix(1, 1, mxREAL);
-  plhs[5] = mxCreateDoubleMatrix(n1, 1, mxCOMPLEX);
-  plhs[6] = mxCreateDoubleMatrix(1, 1, mxREAL);
-
-  /* Assign pointers to the various parameters */
-  double *s = mxGetPr(plhs[1]);
-  double *t = mxGetPr(plhs[2]);
-  double *z = mxGetPr(plhs[3]);
-  double *sdim = mxGetPr(plhs[4]);
-#if MX_HAS_INTERLEAVED_COMPLEX
-  mxComplexDouble *gev = mxGetComplexDoubles(plhs[5]);
-#else
-  double *gev_r = mxGetPr(plhs[5]);
-  double *gev_i = mxGetPi(plhs[5]);
-#endif
-  double *info = mxGetPr(plhs[6]);
-
-  const double *a = mxGetPr(prhs[0]);
-  const double *b = mxGetPr(prhs[1]);
-
-  /* set criterium for stable eigenvalues */
-  if (nrhs >= 3 && mxGetM(prhs[2]) > 0)
-    criterium = *mxGetPr(prhs[2]);
-  else
-    criterium = 1+1e-6;
-
-  /* set criterium for 0/0 generalized eigenvalues */
-  double zhreshold;
-  if (nrhs == 4 && mxGetM(prhs[3]) > 0)
-    zhreshold = *mxGetPr(prhs[3]);
-  else
-    zhreshold = 1e-6;
-
-  /* keep a and b intact */
-  std::copy_n(a, n1*n1, s);
-  std::copy_n(b, n1*n1, t);
-
-  lapack_int i_n = static_cast<lapack_int>(n1);
-  auto alpha_r = std::make_unique<double[]>(n1);
-  auto alpha_i = std::make_unique<double[]>(n1);
-  auto beta = std::make_unique<double[]>(n1);
-  lapack_int lwork = 16*i_n+16;
-  auto work = std::make_unique<double[]>(lwork);
-  auto bwork = std::make_unique<lapack_int[]>(i_n);
-  lapack_int i_info, i_sdim;
-#if defined(MATLAB_MEX_FILE) && MATLAB_VERSION < 0x0904
-  /* The left Schur vectors (VSL) are normally not computed, since JOBVSL="N".
-     But old MKL versions (at least the one shipped with MATLAB R2009b/7.9, which
-     is MKL 10.1) are
-     buggy, and passing nullptr for VSL leads to a crash. Hence we need to
-     allocate space for it.
-     The bug seems to be fixed in MATLAB R2010a/7.10 (MKL 10.2), but we use the
-     workaround for all versions < R2018a/9.4, since those share the same
-     ABI and hence the same executables. */
-  auto vsl = std::make_unique<double[]>(n1*n1);
-#endif
-
-  dgges("N", "V", "S", my_criteria, &i_n, s, &i_n, t, &i_n, &i_sdim, alpha_r.get(), alpha_i.get(),
-        beta.get(),
-#if defined(MATLAB_MEX_FILE) && MATLAB_VERSION < 0x0904
-        vsl.get(),
-#else
-        nullptr,
-#endif
-        &i_n, z, &i_n, work.get(), &lwork, bwork.get(), &i_info);
-
-  *sdim = static_cast<double>(i_sdim);
-  *info = static_cast<double>(i_info);
-
-  for (size_t i = 0; i < n1; i++)
-    {
-      if (std::abs(alpha_r[i]) > zhreshold || std::abs(beta[i]) > zhreshold)
-#if MX_HAS_INTERLEAVED_COMPLEX
-        gev[i].real = alpha_r[i] / beta[i];
-#else
-        gev_r[i] = alpha_r[i] / beta[i];
-#endif
-      else
-        {
-          /* the ratio is too close to 0/0;
-             returns specific error number only if no other error */
-          if (*info == 0)
-            *info = -30;
-        }
-      if (alpha_i[i] == 0.0 && beta[i] == 0.0)
-#if MX_HAS_INTERLEAVED_COMPLEX
-        gev[i].imag = 0.0;
-#else
-        gev_i[i] = 0.0;
-#endif
-      else
-#if MX_HAS_INTERLEAVED_COMPLEX
-        gev[i].imag = alpha_i[i] / beta[i];
-#else
-        gev_i[i] = alpha_i[i] / beta[i];
-#endif
-    }
-
-  plhs[0] = mxCreateDoubleScalar(0);
-}