Work-In-Progress commit for new ELPA interface

Makefile.am

@@ -20,6 +20,7 @@ libelpa@SUFFIX@_public_la_SOURCES = \
 	src/elpa1/legacy_interface/elpa1_auxiliary_legacy.F90 \
 	src/elpa1/elpa1_auxiliary.F90 \
 	src/elpa_t.F90 \
+	src/elpa_constants.F90 \
         src/general/elpa_utilities.F90
 
 # internal parts
@@ -53,8 +54,8 @@ libelpa@SUFFIX@_private_la_SOURCES = \
         src/elpa2/qr/elpa_pdgeqrf.F90 \
 	src/elpa1/elpa1.F90 \
 	src/elpa2/elpa2.F90 \
-	src/elpa_index.c \
-	src/elpa_options.c
+	src/elpa_constants.c \
+	src/elpa_index.c
 
 EXTRA_libelpa@SUFFIX@_private_la_DEPENDENCIES = \
         src/elpa1/elpa_reduce_add_vectors.X90 \
@@ -322,10 +323,9 @@ endif
 include generated_headers.am
 BUILT_SOURCES = $(generated_headers)
 
-# install public Fortran modules files in the include/ dir
+# install public headers and Fortran modules files in the include/ dir
 elpa_includedir = $(includedir)/elpa@SUFFIX@-@PACKAGE_VERSION@
-nobase_elpa_include_HEADERS = $(wildcard modules/*)
-nobase_elpa_include_HEADERS += elpa/elpa.h elpa/elpa_kernel_constants.h elpa/elpa_solver_constants.h elpa/elpa_constants.h elpa/elpa_generated.h
+nobase_elpa_include_HEADERS = $(wildcard modules/*) $(wildcard elpa/*)
 
 dist_man_MANS = \
            man/solve_evp_real.3 \

elpa/elpa_constants.h

-
-#define ELPA_INVALID_INT INT_MIN
-
-#define ELPA_C_ERROR    0
-#define ELPA_C_OK       1
-
-#ifdef ELPA_H
-
-#define ELPA_ERROR      ELPA_C_ERROR
-#define ELPA_OK         ELPA_C_OK
-
-#endif
-
-#include <elpa/elpa_kernel_constants.h>
-#include <elpa/elpa_solver_constants.h>

elpa/elpa_constants.h.in

+#pragma once
+
+/* This might seem over-engineered, but helps to re-use this file also on the
+ * Fortran side and thus to keep the definitions in this one place here
+ */
+
+/* Private helper macros */
+#define ELPA_ENUM_ENTRY(name, value) \
+        name = value,
+#define ELPA_ENUM_SUM(name, value) +1
+
+
+/* Solver constants */
+#define ELPA_SOLVER_MAP(X) \
+        X(ELPA_SOLVER_1STAGE, 1) \
+        X(ELPA_SOLVER_2STAGE, 2)
+
+enum ELPA_SOLVERS {
+        ELPA_SOLVER_MAP(ELPA_ENUM_ENTRY)
+};
+
+#define ELPA_NUMBER_OF_SOLVERS (0 ELPA_SOLVER_MAP(ELPA_ENUM_SUM))
+
+/* Kernel constants */
+#define ELPA_REAL_KERNEL_MAP(X) \
+        X(ELPA_2STAGE_REAL_GENERIC, 1) \
+        X(ELPA_2STAGE_REAL_GENERIC_SIMPLE, 2) \
+        X(ELPA_2STAGE_REAL_BGP, 3) \
+        X(ELPA_2STAGE_REAL_BGQ, 4) \
+        X(ELPA_2STAGE_REAL_SSE, 5) \
+        X(ELPA_2STAGE_REAL_SSE_BLOCK2, 6) \
+        X(ELPA_2STAGE_REAL_SSE_BLOCK4, 7) \
+        X(ELPA_2STAGE_REAL_SSE_BLOCK6, 8) \
+        X(ELPA_2STAGE_REAL_AVX_BLOCK2, 9) \
+        X(ELPA_2STAGE_REAL_AVX_BLOCK4, 10) \
+        X(ELPA_2STAGE_REAL_AVX_BLOCK6, 11) \
+        X(ELPA_2STAGE_REAL_AVX2_BLOCK2, 12) \
+        X(ELPA_2STAGE_REAL_AVX2_BLOCK4, 13) \
+        X(ELPA_2STAGE_REAL_AVX2_BLOCK6, 14) \
+        X(ELPA_2STAGE_REAL_AVX512_BLOCK2, 15) \
+        X(ELPA_2STAGE_REAL_AVX512_BLOCK4, 16) \
+        X(ELPA_2STAGE_REAL_AVX512_BLOCK6, 17) \
+        X(ELPA_2STAGE_REAL_GPU, 18) \
+        X(ELPA_2STAGE_REAL_DEFAULT, @ELPA_2STAGE_REAL_DEFAULT@)
+
+enum ELPA_REAL_KERNELS {
+        ELPA_REAL_KERNEL_MAP(ELPA_ENUM_ENTRY)
+};
+
+
+#define ELPA_COMPLEX_KERNEL_MAP(X) \
+        X(ELPA_2STAGE_COMPLEX_GENERIC, 1) \
+        X(ELPA_2STAGE_COMPLEX_GENERIC_SIMPLE, 2) \
+        X(ELPA_2STAGE_COMPLEX_BGP, 3) \
+        X(ELPA_2STAGE_COMPLEX_BGQ, 4) \
+        X(ELPA_2STAGE_COMPLEX_SSE, 5) \
+        X(ELPA_2STAGE_COMPLEX_SSE_BLOCK1, 6) \
+        X(ELPA_2STAGE_COMPLEX_SSE_BLOCK2, 7) \
+        X(ELPA_2STAGE_COMPLEX_AVX_BLOCK1, 8) \
+        X(ELPA_2STAGE_COMPLEX_AVX_BLOCK2, 9) \
+        X(ELPA_2STAGE_COMPLEX_AVX2_BLOCK1, 10) \
+        X(ELPA_2STAGE_COMPLEX_AVX2_BLOCK2, 11) \
+        X(ELPA_2STAGE_COMPLEX_AVX512_BLOCK1, 12) \
+        X(ELPA_2STAGE_COMPLEX_AVX512_BLOCK2, 13) \
+        X(ELPA_2STAGE_COMPLEX_GPU, 14) \
+        X(ELPA_2STAGE_COMPLEX_DEFAULT, @ELPA_2STAGE_COMPLEX_DEFAULT@)
+
+enum ELPA_COMPLEX_KERNELS {
+        ELPA_COMPLEX_KERNEL_MAP(ELPA_ENUM_ENTRY)
+};
+
+
+
+/* General constants */
+#define ELPA_CONSTANTS_MAP(X) \
+        X(ELPA_ERROR, -1) \
+        X(ELPA_NO, 0) \
+        X(ELPA_OK, 1)
+
+enum ELPA_CONSTANTS {
+        ELPA_CONSTANTS_MAP(ELPA_ENUM_ENTRY)
+
+        /* -1 to take out the default kernel again */
+        ELPA_2STAGE_NUMBER_OF_COMPLEX_KERNELS = (0 ELPA_COMPLEX_KERNEL_MAP(ELPA_ENUM_SUM) -1),
+        ELPA_2STAGE_NUMBER_OF_REAL_KERNELS = (0 ELPA_REAL_KERNEL_MAP(ELPA_ENUM_SUM) -1),
+};
+

elpa/elpa_solver_constants.h

-#define ELPA_C_SOLVER_1STAGE 1
-#define ELPA_C_SOLVER_2STAGE 2
-#define ELPA_C_NUMBER_OF_SOLVERS 2
-
-#ifdef ELPA_H
-
-#define ELPA_SOLVER_1STAGE     ELPA_C_SOLVER_1STAGE
-#define ELPA_SOLVER_2STAGE     ELPA_C_SOLVER_2STAGE
-#define ELPA_NUMBER_OF_SOLVERS ELPA_C_NUMBER_OF_SOLVERS
-
-#endif

generated_headers.am

@@ -28,7 +28,13 @@ elpa/elpa_generated_fortran_interfaces.h: $(wildcard $(top_srcdir)/src/elpa2/ker
 	$(call extract_interface,!f>)
 	$(call extract_interface,#!f>)
 
-generated_headers= config-f90.h elpa/elpa_generated.h test/shared/generated.h elpa/elpa_generated_fortran_interfaces.h
+src/fortran_constants.X90: $(top_srcdir)/src/fortran_constants.h
+	$(CPP) $(CPPFLAGS) -I$(top_builddir)/ -I$(top_srcdir)/ -I. $< -o $@_ || { rm -f $@; exit 1; }
+	awk '/!ELPA_C_DEFINE/ {gsub(/!ELPA_C_DEFINE/, "\n"); gsub(/NEWLINE/, "\n"); print;}' < $@_ > $@ || { rm -f $@; exit 1; }
+	rm $@_
+
+
+generated_headers= config-f90.h elpa/elpa_generated.h test/shared/generated.h elpa/elpa_generated_fortran_interfaces.h src/fortran_constants.X90
 generated-headers: $(generated_headers)
 
 

m4/ax_elpa_gpu_version_only.m4

-
-dnl macro for testing whether the user wanted to compile only with the GPU version
-
-dnl usage: DEFINE_OPTION([gpu-support-only],[gpu-support],[with_gpu_support],[install_gpu])
-
-AC_DEFUN([DEFINE_OPTION_GPU_SUPPORT_ONLY],[
-  AC_ARG_WITH([$1],
-               AS_HELP_STRING([--with-$1],
-                              [only compile $2 ]),
-              [with_option=yes],[with_option=no])
-
-  if test x"${with_option}" = x"yes" ; then
-    dnl make sure that all the other kernels are unset
-    install_real_generic=no
-    install_real_generic_simple=no
-    install_real_sse=no
-    install_real_bgp=no
-    install_real_bgq=no
-    install_real_avx_block2=no
-    install_real_avx_block4=no
-    install_real_avx_block6=no
-
-
-    install_complex_generic=no
-    install_complex_generic_simple=no
-    install_complex_sse=no
-    install_complex_bgp=no
-    install_complex_bgq=no
-    install_complex_avx_block1=no
-    install_complex_avx_block2=no
-
-
-    install_gpu=yes
-
-    want_avx=no
-
-    build_with_gpu_support_only=yes
-    use_specific_complex_kernel=yes
-    use_specific_real_kernel=yes
-    dnl now set the specific kernel
-    $3=yes
-
-    AC_MSG_NOTICE([ELPA will be build only with $1])
-  else
-    build_with_gpu_support_only=no
-  fi
-])
-

m4/ax_prog_cxx_mpi.m4

-# ===========================================================================
-#      http://www.gnu.org/software/autoconf-archive/ax_prog_cxx_mpi.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_CXX_MPI([MPI-WANTED-TEST[, ACTION-IF-FOUND[, ACTION-IF-NOT-FOUND]]])
-#
-# DESCRIPTION
-#
-#   This macro tries to find out how to compile C++ programs that use MPI
-#   (Message Passing Interface), a standard API for parallel process
-#   communication (see http://www-unix.mcs.anl.gov/mpi/).  The macro has to
-#   be used instead of the standard macro AC_PROG_CXX and will replace the
-#   standard variable CXX with the found compiler.
-#
-#   MPI-WANTED-TEST is used to test whether MPI is actually wanted by the
-#   user. If MPI-WANTED_TEST is omitted or if it succeeds, the macro will
-#   try to find out how to use MPI, if it fails, the macro will call
-#   AC_PROG_CC to find a standard C compiler instead.
-#
-#   When MPI is found, ACTION-IF-FOUND will be executed, if MPI is not found
-#   (or MPI-WANTED-TEST fails) ACTION-IF-NOT-FOUND is executed. If
-#   ACTION-IF-FOUND is not set, the macro will define HAVE_MPI.
-#
-#   The following example demonstrates usage of the macro:
-#
-#     # If --with-mpi=auto is used, try to find MPI, but use standard C compiler if it is not found.
-#     # If --with-mpi=yes is used, try to find MPI and fail if it isn't found.
-#     # If --with-mpi=no is used, use a standard C compiler instead.
-#     AC_ARG_WITH(mpi, [AS_HELP_STRING([--with-mpi],
-#         [compile with MPI (parallelization) support. If none is found,
-#         MPI is not used. Default: auto])
-#     ],,[with_mpi=auto])
-#
-#     AX_PROG_CXX_MPI([test x"$with_mpi" != xno],[use_mpi=yes],[
-#       use_mpi=no
-#       if test x"$with_mpi" = xyes; then
-#         AC_MSG_FAILURE([MPI compiler requested, but couldn't use MPI.])
-#       else
-#         AC_MSG_WARN([No MPI compiler found, won't use MPI.])
-#       fi
-#     ])
-#
-# LICENSE
-#
-#   Copyright (c) 2010,2011 Olaf Lenz <olenz@icp.uni-stuttgart.de>
-#
-#   This program 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.
-#
-#   This program 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 this program. If not, see <http://www.gnu.org/licenses/>.
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 2
-
-AC_DEFUN([AX_PROG_CXX_MPI], [
-AC_PREREQ(2.50)
-
-# Check for compiler
-# Needs to be split off into an extra macro to ensure right expansion
-# order.
-AC_REQUIRE([_AX_PROG_CXX_MPI],[_AX_PROG_CXX_MPI([$1])])
-
-AS_IF([test x"$_ax_prog_cxx_mpi_mpi_wanted" = xno],
-  [ _ax_prog_cxx_mpi_mpi_found=no ],
-  [
-    AC_LANG_PUSH([C++])
-
-    # test whether MPI_Init() is available
-    # We do not use AC_SEARCH_LIBS here, as it caches its outcome and
-    # thus disallows corresponding calls in the other AX_PROG_*_MPI
-    # macros.
-    for lib in NONE mpi mpich; do
-      save_LIBS=$LIBS
-      if test x"$lib" = xNONE; then
-        AC_MSG_CHECKING([for function MPI_Init])
-      else
-        AC_MSG_CHECKING([for function MPI_Init in -l$lib])
-        LIBS="-l$lib $LIBS"
-      fi
-      AC_LINK_IFELSE([
-        AC_LANG_PROGRAM([
-extern "C" { void MPI_Init(); }
-],[MPI_Init();])],
-        [ _ax_prog_cxx_mpi_mpi_found=yes ],
-        [ _ax_prog_cxx_mpi_mpi_found=no ])
-      AC_MSG_RESULT($_ax_prog_cxx_mpi_mpi_found)
-      if test "x$_ax_prog_cxx_mpi_mpi_found" = "xyes"; then
-        break;
-      fi
-      LIBS=$save_LIBS
-    done
-
-    # Check for header
-    AS_IF([test x"$_ax_prog_cxx_mpi_mpi_found" = xyes], [
-      AC_MSG_CHECKING([for mpi.h])
-      AC_COMPILE_IFELSE([AC_LANG_PROGRAM([#include <mpi.h>])],
-        [ AC_MSG_RESULT(yes)],
-        [ AC_MSG_RESULT(no)
-         _ax_prog_cxx_mpi_mpi_found=no
-      ])
-    ])
-    AC_LANG_POP([C++])
-])
-
-# Finally, execute ACTION-IF-FOUND/ACTION-IF-NOT-FOUND:
-AS_IF([test x"$_ax_prog_cxx_mpi_mpi_found" = xyes], [
-        ifelse([$2],,[AC_DEFINE(HAVE_MPI,1,[Define if you have the MPI library.])],[$2])
-        :
-],[
-        $3
-        :
-])
-
-])dnl AX_PROG_CXX_MPI
-
-dnl _AX_PROG_CXX_MPI is an internal macro required by AX_PROG_CXX_MPI.
-dnl To ensure the right expansion order, the main function AX_PROG_CXX_MPI
-dnl has to be split into two parts.
-dnl
-dnl Known MPI C++ compilers:
-dnl  mpic++
-dnl  mpicxx
-dnl  mpiCC
-dnl  sxmpic++     NEC SX
-dnl  hcp
-dnl  mpxlC_r
-dnl  mpxlC
-dnl  mpixlcxx_r
-dnl  mpixlcxx
-dnl  mpg++
-dnl  mpc++
-dnl  mpCC
-dnl  cmpic++
-dnl  mpiFCC       Fujitsu
-dnl  CC
-dnl
-AC_DEFUN([_AX_PROG_CXX_MPI], [
-  ifelse([$1],,[_ax_prog_cxx_mpi_mpi_wanted=yes],[
-    AC_MSG_CHECKING([whether to compile using MPI])
-    if $1; then
-      _ax_prog_cxx_mpi_mpi_wanted=yes
-    else
-      _ax_prog_cxx_mpi_mpi_wanted=no
-    fi
-    AC_MSG_RESULT($_ax_prog_cxx_mpi_mpi_wanted)
-  ])
-  if test x"$_ax_prog_cxx_mpi_mpi_wanted" = xyes; then
-    AC_CHECK_TOOLS([CXX], [mpic++ mpicxx mpiCC sxmpic++ hcp mpxlC_r mpxlC mpixlcxx_r mpixlcxx mpg++ mpc++ mpCC cmpic++ mpiFCC CCicpc pgCC pathCC sxc++ xlC_r xlC bgxlC_r bgxlC openCC sunCC crayCC g++ c++ gpp aCC CC cxx cc++ cl.exe FCC KCC RCC])
-  fi
-  AC_PROG_CXX
-])dnl _AX_PROG_CXX_MPI

src/elpa1/legacy_interface/elpa_cholesky_template_legacy.X90

@@ -89,7 +89,7 @@
 !      integer(kind=ik)              :: istat
 !      character(200)                :: errorMessage
 
-      type(elpa_t)                 :: elpaAPI
+      type(elpa_t)                 :: e
 
       call timer%start("elpa_cholesky_&
       &MATH_DATATYPE&
@@ -106,8 +106,18 @@
         return
       endif
 
-      nev = 10
-      elpaAPI = elpa_create(na, nev, lda, matrixCols, nblk, successInternal)
+      e = elpa_allocate()
+
+      call e%set("na", na)
+      call e%set("local_nrows", lda)
+      call e%set("local_ncols", matrixCols)
+      call e%set("nblk", nblk)
+
+      call e%set("mpi_comm_rows", mpi_comm_rows)
+      call e%set("mpi_comm_cols", mpi_comm_cols)
+
+      call e%setup(successInternal)
+
       if (successInternal .ne. ELPA_OK) then
         print *, "Cannot create elpa object"
         success = .false.
@@ -115,10 +125,8 @@
         return
       endif
 
-      call elpaAPI%set_comm_rows(mpi_comm_rows)
-      call elpaAPI%set_comm_cols(mpi_comm_cols)
 
-      call elpaAPI%cholesky(a(1:lda,1:matrixCols), successInternal)
+      call e%cholesky(a(1:lda,1:matrixCols), successInternal)
 
       if (successInternal .ne. ELPA_OK) then
         print *, "Cannot run cholesky"
@@ -128,7 +136,7 @@
       else
         success =.true.
       endif
-      call elpaAPI%destroy()
+      call e%destroy()
 
       call elpa_uninit()
 

src/elpa1/legacy_interface/elpa_invert_trm_legacy.X90

@@ -81,9 +81,9 @@
 #endif
 #endif
 
- !      integer(kind=ik)             :: my_prow, my_pcol, np_rows, np_cols, mpierr
- !      integer(kind=ik)             :: l_cols, l_rows, l_col1, l_row1, l_colx, l_rowx
- !      integer(kind=ik)             :: n, nc, i, info, ns, nb
+!      integer(kind=ik)             :: my_prow, my_pcol, np_rows, np_cols, mpierr
+!      integer(kind=ik)             :: l_cols, l_rows, l_col1, l_row1, l_colx, l_rowx
+!      integer(kind=ik)             :: n, nc, i, info, ns, nb
 !#if REALCASE ==  1
 !       real(kind=REAL_DATATYPE), allocatable   :: tmp1(:), tmp2(:,:), tmat1(:,:), tmat2(:,:)
 !#endif
@@ -92,12 +92,12 @@
 !#endif
        logical, intent(in)          :: wantDebug
        logical                      :: success
- !      integer(kind=ik)             :: istat
- !      character(200)               :: errorMessage
+!      integer(kind=ik)             :: istat
+!      character(200)               :: errorMessage
 
        integer(kind=ik)            :: nev
        integer(kind=iK)            :: successInternal
-       type(elpa_t)                :: elpaAPI
+       type(elpa_t)                :: e
 
        call timer%start("elpa_invert_trm_&
        &MATH_DATATYPE&
@@ -114,8 +114,18 @@
          return
        endif
 
-       nev = 10
-       elpaAPI = elpa_create(na, nev, lda, matrixCols, nblk, successInternal)
+       e = elpa_allocate()
+
+       call e%set("na", na)
+       call e%set("local_nrows", lda)
+       call e%set("local_ncols", matrixCols)
+       call e%set("nblk", nblk)
+
+       call e%set("mpi_comm_rows", mpi_comm_rows)
+       call e%set("mpi_comm_cols", mpi_comm_cols)
+
+       call e%setup(successInternal)
+
        if (successInternal .ne. ELPA_OK) then
          print *, "Cannot create elpa object"
          success = .false.
@@ -123,10 +133,8 @@
          return
        endif
 
-       call elpaAPI%set_comm_rows(mpi_comm_rows)
-       call elpaAPI%set_comm_cols(mpi_comm_cols)
 
-       call elpaAPI%invert_trm(a(1:lda,1:matrixCols), successInternal)
+       call e%invert_tridiagonal(a(1:lda,1:matrixCols), successInternal)
 
        if (successInternal .ne. ELPA_OK) then
          print *, "Cannot run invert_trm"
@@ -136,7 +144,7 @@
        else
          success =.true.
        endif
-       call elpaAPI%destroy()
+       call e%destroy()
 
        call elpa_uninit()
 

src/elpa1/legacy_interface/elpa_multiply_a_b_legacy.X90

@@ -102,7 +102,7 @@
 !      character(200)                              :: errorMessage
       logical                                     :: success
       integer(kind=ik)                            :: successInternal
-      type(elpa_t)                                :: elpaAPI
+      type(elpa_t)                                :: e
 
       call timer%start("elpa_mult_at_b_&
       &MATH_DATATYPE&
@@ -127,19 +127,25 @@
         return
       endif
 
-      nev = 10
-      elpaAPI = elpa_create(na, nev, lda, ldaCols, nblk, successInternal)
+      e = elpa_allocate()
+
+      call e%set("na", na)
+      call e%set("local_nrows", lda)
+      call e%set("local_ncols", matrixCols)
+      call e%set("nblk", nblk)
+
+      call e%set("mpi_comm_rows", mpi_comm_rows)
+      call e%set("mpi_comm_cols", mpi_comm_cols)
+
+      call e%setup(successInternal)
       if (successInternal .ne. ELPA_OK) then
-        print *, "Cannot create elpa object"
-        success = .false.
+        print *, "Cannot run multiply_a_b"
         stop
+        success = .false.
         return
       endif
 
-      call elpaAPI%set_comm_rows(mpi_comm_rows)
-      call elpaAPI%set_comm_cols(mpi_comm_cols)
-
-      call elpaAPI%multiply_a_b(uplo_a, uplo_c, na, ncb, a(1:lda,1:ldaCols), lda, ldaCols, &
+      call e%hermitian_multiply(uplo_a, uplo_c, na, ncb, a(1:lda,1:ldaCols), lda, ldaCols, &
                                 b(1:ldb,1:ldbCols), ldb, ldbCols, &
                                 c(1:ldc,1:ldcCols), ldc, ldcCols, successInternal)
 
@@ -149,7 +155,7 @@
         success = .false.
         return
       endif
-      call elpaAPI%destroy()
+      call e%destroy()
 
       call elpa_uninit()
 

src/elpa2/kernels/elpa2_kernels_complex_sse_1hv_double_precision.c

@@ -78,7 +78,7 @@ static __forceinline void hh_trafo_complex_kernel_4_SSE_1hv_double(double comple
 static __forceinline void hh_trafo_complex_kernel_2_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
 
 /*
-!f>#ifdef HAVE_SSE_INTRINSICS
+!f>#ifdef WITH_COMPLEX_SSE_BLOCK1_KERNEL
 !f> interface
 !f>   subroutine single_hh_trafo_complex_sse_1hv_double(q, hh, pnb, pnq, pldq) &
 !f>                             bind(C, name="single_hh_trafo_complex_sse_1hv_double")

src/elpa2/legacy_interface/elpa2_utilities_legacy.F90

@@ -49,17 +49,15 @@
 !
 ! Author: Andreas Marek, MPCDF
 
-
 #include "config-f90.h"
-#include <elpa/elpa_kernel_constants.h>
 
 module elpa2_utilities
-  !use elpa2_utilities_private
+  use elpa_constants
   use elpa_utilities
   use precision
   implicit none
 
-  PRIVATE ! By default, all routines contained are private
+  private ! by default, all routines contained are private
 
   ! The following routines are public:
 
@@ -94,26 +92,26 @@ module elpa2_utilities
 
   public :: elpa_real_kernel_name, elpa_complex_kernel_name
 
-  integer(kind=ik), parameter :: number_of_real_kernels           = ELPA_C_2STAGE_NUMBER_OF_REAL_KERNELS
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC         = ELPA_C_2STAGE_REAL_GENERIC
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC_SIMPLE  = ELPA_C_2STAGE_REAL_GENERIC_SIMPLE
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGP             = ELPA_C_2STAGE_REAL_BGP
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGQ             = ELPA_C_2STAGE_REAL_BGQ
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE             = ELPA_C_2STAGE_REAL_SSE
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE_BLOCK2      = ELPA_C_2STAGE_REAL_SSE_BLOCK2
-  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE_BLOCK4      = ELPA_C_2STAGE_REAL_SSE_BLOCK4
+  integer(kind=ik), parameter :: number_of_real_kernels = ELPA_2STAGE_NUMBER_OF_REAL_KERNELS
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC = ELPA_2STAGE_REAL_GENERIC
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC_SIMPLE = ELPA_2STAGE_REAL_GENERIC_SIMPLE
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGP = ELPA_2STAGE_REAL_BGP
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGQ = ELPA_2STAGE_REAL_BGQ
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE = ELPA_2STAGE_REAL_SSE
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE_BLOCK2 = ELPA_2STAGE_REAL_SSE_BLOCK2
+  integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE_BLOCK4 = ELPA_2STAGE_R