Commit 24867b0e authored by Andreas Marek's avatar Andreas Marek

Merge branch 'master' into ELPA_GPU

parents 3896c305 ebc097eb
...@@ -95,24 +95,44 @@ endif ...@@ -95,24 +95,44 @@ endif
endif endif
endif endif
if WITH_REAL_SSE_BLOCK2_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse_2hv.c
endif
if WITH_REAL_AVX_BLOCK2_KERNEL if WITH_REAL_AVX_BLOCK2_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_avx-avx2_2hv.c
endif
if WITH_REAL_SSE_BLOCK4_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse_4hv.c
endif endif
if WITH_REAL_AVX_BLOCK4_KERNEL if WITH_REAL_AVX_BLOCK4_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_avx-avx2_4hv.c
endif
if WITH_REAL_SSE_BLOCK6_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse_6hv.c
endif endif
if WITH_REAL_AVX_BLOCK6_KERNEL if WITH_REAL_AVX_BLOCK6_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_avx-avx2_6hv.c
endif
if WITH_COMPLEX_SSE_BLOCK1_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_sse_1hv.cpp
endif endif
if WITH_COMPLEX_AVX_BLOCK1_KERNEL if WITH_COMPLEX_AVX_BLOCK1_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_sse-avx_1hv.cpp libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_avx-avx2_1hv.cpp
endif
if WITH_COMPLEX_SSE_BLOCK2_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_sse_2hv.cpp
endif endif
if WITH_COMPLEX_AVX_BLOCK2_KERNEL if WITH_COMPLEX_AVX_BLOCK2_KERNEL
libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_sse-avx_2hv.cpp libelpa@SUFFIX@_la_SOURCES += src/elpa2_kernels/elpa2_kernels_complex_avx-avx2_2hv.cpp
endif endif
.cu.lo: .cu.lo:
......
...@@ -151,6 +151,10 @@ install_real_generic_simple=yes ...@@ -151,6 +151,10 @@ install_real_generic_simple=yes
install_complex_generic=yes install_complex_generic=yes
install_complex_generic_simple=yes install_complex_generic_simple=yes
#want_avx=yes
#want_avx2=yes
#want_sse=yes
AC_LANG([C]) AC_LANG([C])
dnl build with ftimings support dnl build with ftimings support
...@@ -204,12 +208,26 @@ $CC -c $srcdir/src/elpa2_kernels/elpa2_kernels_asm_x86_64_double_precision.s -o ...@@ -204,12 +208,26 @@ $CC -c $srcdir/src/elpa2_kernels/elpa2_kernels_asm_x86_64_double_precision.s -o
if test "$?" == 0; then if test "$?" == 0; then
can_compile_sse=yes can_compile_sse=yes
install_real_sse=yes install_real_sse=yes
install_real_sse_block2=yes
install_real_sse_block4=yes
install_real_sse_block6=yes
install_complex_sse=yes install_complex_sse=yes
install_complex_sse_block1=yes
install_complex_sse_block2=yes
else else
can_compile_sse=no can_compile_sse=no
install_real_sse=no install_real_sse=no
install_real_sse_block2=no
install_real_sse_block4=no
install_real_sse_block6=no
install_complex_sse=no install_complex_sse=no
install_complex_sse_block1=no
install_complex_sse_block2=no
fi fi
rm -f ./test.o rm -f ./test.o
AC_MSG_RESULT([${can_compile_sse}]) AC_MSG_RESULT([${can_compile_sse}])
...@@ -249,6 +267,7 @@ AC_COMPILE_IFELSE([AC_LANG_SOURCE([ ...@@ -249,6 +267,7 @@ AC_COMPILE_IFELSE([AC_LANG_SOURCE([
[can_compile_avx=no] [can_compile_avx=no]
) )
AC_MSG_RESULT([${can_compile_avx}]) AC_MSG_RESULT([${can_compile_avx}])
if test "${can_compile_avx}" = "yes" ; then if test "${can_compile_avx}" = "yes" ; then
AC_MSG_CHECKING([whether we can compile AVX intrinsics in C++]) AC_MSG_CHECKING([whether we can compile AVX intrinsics in C++])
AC_LANG_PUSH([C++]) AC_LANG_PUSH([C++])
...@@ -306,7 +325,6 @@ if test "${can_compile_avx2}" = "yes" ; then ...@@ -306,7 +325,6 @@ if test "${can_compile_avx2}" = "yes" ; then
fi fi
fi fi
if test "${can_compile_avx}" = "yes" ; then if test "${can_compile_avx}" = "yes" ; then
install_real_avx_block2=yes install_real_avx_block2=yes
install_real_avx_block4=yes install_real_avx_block4=yes
...@@ -314,8 +332,6 @@ if test "${can_compile_avx}" = "yes" ; then ...@@ -314,8 +332,6 @@ if test "${can_compile_avx}" = "yes" ; then
install_complex_avx_block1=yes install_complex_avx_block1=yes
install_complex_avx_block2=yes install_complex_avx_block2=yes
want_avx=yes
else else
install_real_avx_block2=no install_real_avx_block2=no
install_real_avx_block4=no install_real_avx_block4=no
...@@ -323,8 +339,34 @@ else ...@@ -323,8 +339,34 @@ else
install_complex_avx_block1=no install_complex_avx_block1=no
install_complex_avx_block2=no install_complex_avx_block2=no
fi
if test "${can_compile_avx2}" = "yes" ; then
install_real_avx2_block2=yes
install_real_avx2_block4=yes
install_real_avx2_block6=yes
want_avx=yes install_complex_avx2_block1=yes
install_complex_avx2_block2=yes
else
install_real_avx2_block2=no
install_real_avx2_block4=no
install_real_avx2_block6=no
install_complex_avx2_block1=no
install_complex_avx2_block2=no
fi
AM_CONDITIONAL([HAVE_SSE],[test x"$can_compile_sse" = x"yes"])
if test x"${can_compile_sse}" = x"yes" ; then
AC_DEFINE([HAVE_SSE],[1],[SSE is supported on this CPU])
fi
AM_CONDITIONAL([HAVE_AVX],[test x"$can_compile_avx" = x"yes"])
if test x"${can_compile_avx}" = x"yes" ; then
AC_DEFINE([HAVE_AVX],[1],[AVX is supported on this CPU])
fi
AM_CONDITIONAL([HAVE_AVX2],[test x"$can_compile_avx2" = x"yes"])
if test x"${can_compile_avx2}" = x"yes" ; then
AC_DEFINE([HAVE_AVX2],[1],[AVX2 is supported on this CPU])
fi fi
dnl set the AVX optimization flags if this option is specified dnl set the AVX optimization flags if this option is specified
...@@ -492,7 +534,6 @@ AC_MSG_CHECKING([whether we can use the intrinsic Fortran function "get_environm ...@@ -492,7 +534,6 @@ AC_MSG_CHECKING([whether we can use the intrinsic Fortran function "get_environm
AC_COMPILE_IFELSE([AC_LANG_SOURCE([ AC_COMPILE_IFELSE([AC_LANG_SOURCE([
program test_get_environment program test_get_environment
character(len=256) :: homedir character(len=256) :: homedir
call get_environment_variable("HOME",homedir) call get_environment_variable("HOME",homedir)
end program end program
...@@ -638,6 +679,15 @@ DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-bgp-kernel-only],[bgp-kernel],[install_ ...@@ -638,6 +679,15 @@ DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-bgp-kernel-only],[bgp-kernel],[install_
dnl bgq kernel dnl bgq kernel
DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-bgq-kernel-only],[bgq-kernel],[install_real_bgq]) DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-bgq-kernel-only],[bgq-kernel],[install_real_bgq])
dnl real-sse-block2 kernel
DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-sse-block2-kernel-only],[real-sse-block2-kernel],[install_real_sse_block2])
dnl real-sse-block4 kernel
DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-sse-block4-kernel]-only,[real-sse-block4-kernel],[install_real_sse_block4])
dnl real-sse-block6 kernel
DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-sse-block6-kernel-only],[real-sse-block6-kernel],[install_real_sse_block6])
dnl real-avx-block2 kernel dnl real-avx-block2 kernel
DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-avx-block2-kernel-only],[real-avx-block2-kernel],[install_real_avx_block2]) DEFINE_OPTION_SPECIFIC_REAL_KERNEL([real-avx-block2-kernel-only],[real-avx-block2-kernel],[install_real_avx_block2])
...@@ -664,6 +714,12 @@ DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-bgp-kernel-only],[bgp-kernel],[in ...@@ -664,6 +714,12 @@ DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-bgp-kernel-only],[bgp-kernel],[in
dnl complex-bqq kernel dnl complex-bqq kernel
DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-bgq-kernel-only],[bgq-kernel],[install_complex_bgq]) DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-bgq-kernel-only],[bgq-kernel],[install_complex_bgq])
dnl complex-sse-block1 kernel
DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-sse-block1-kernel-only],[complex-sse-block1-kernel],[install_complex_sse_block1])
dnl complex-sse-block2 kernel
DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-sse-block2-kernel-only],[complex-sse-block2-kernel],[install_complex_sse_block2])
dnl complex-avx-block1 kernel dnl complex-avx-block1 kernel
DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-avx-block1-kernel-only],[complex-avx-block1-kernel],[install_complex_avx_block1]) DEFINE_OPTION_SPECIFIC_COMPLEX_KERNEL([complex-avx-block1-kernel-only],[complex-avx-block1-kernel],[install_complex_avx_block1])
...@@ -711,6 +767,21 @@ if test x"${install_complex_sse}" = x"yes" ; then ...@@ -711,6 +767,21 @@ if test x"${install_complex_sse}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_SSE_KERNEL],[1],[can use complex SSE kernel]) AC_DEFINE([WITH_COMPLEX_SSE_KERNEL],[1],[can use complex SSE kernel])
fi fi
AM_CONDITIONAL([WITH_REAL_SSE_BLOCK2_KERNEL],[test x"$install_real_sse_block2" = x"yes"])
if test x"${install_real_sse_block2}" = x"yes" ; then
AC_DEFINE([WITH_REAL_SSE_BLOCK2_KERNEL],[1],[can use real_sse_block2 kernel])
fi
AM_CONDITIONAL([WITH_REAL_SSE_BLOCK4_KERNEL],[test x"$install_real_sse_block4" = x"yes"])
if test x"${install_real_sse_block4}" = x"yes" ; then
AC_DEFINE([WITH_REAL_SSE_BLOCK4_KERNEL],[1],[can use real_sse_block4 kernel])
fi
AM_CONDITIONAL([WITH_REAL_SSE_BLOCK6_KERNEL],[test x"$install_real_sse_block6" = x"yes"])
if test x"${install_real_sse_block6}" = x"yes" ; then
AC_DEFINE([WITH_REAL_SSE_BLOCK6_KERNEL],[1],[can use real_sse_block6 kernel])
fi
AM_CONDITIONAL([WITH_REAL_AVX_BLOCK2_KERNEL],[test x"$install_real_avx_block2" = x"yes"]) AM_CONDITIONAL([WITH_REAL_AVX_BLOCK2_KERNEL],[test x"$install_real_avx_block2" = x"yes"])
if test x"${install_real_avx_block2}" = x"yes" ; then if test x"${install_real_avx_block2}" = x"yes" ; then
AC_DEFINE([WITH_REAL_AVX_BLOCK2_KERNEL],[1],[can use real_avx_block2 kernel]) AC_DEFINE([WITH_REAL_AVX_BLOCK2_KERNEL],[1],[can use real_avx_block2 kernel])
...@@ -726,6 +797,31 @@ if test x"${install_real_avx_block6}" = x"yes" ; then ...@@ -726,6 +797,31 @@ if test x"${install_real_avx_block6}" = x"yes" ; then
AC_DEFINE([WITH_REAL_AVX_BLOCK6_KERNEL],[1],[can use real_avx_block6 kernel]) AC_DEFINE([WITH_REAL_AVX_BLOCK6_KERNEL],[1],[can use real_avx_block6 kernel])
fi fi
AM_CONDITIONAL([WITH_REAL_AVX2_BLOCK2_KERNEL],[test x"$install_real_avx2_block2" = x"yes"])
if test x"${install_real_avx2_block2}" = x"yes" ; then
AC_DEFINE([WITH_REAL_AVX2_BLOCK2_KERNEL],[1],[can use real_avx2_block2 kernel])
fi
AM_CONDITIONAL([WITH_REAL_AVX2_BLOCK4_KERNEL],[test x"$install_real_avx2_block4" = x"yes"])
if test x"${install_real_avx2_block4}" = x"yes" ; then
AC_DEFINE([WITH_REAL_AVX2_BLOCK4_KERNEL],[1],[can use real_avx2_block4 kernel])
fi
AM_CONDITIONAL([WITH_REAL_AVX2_BLOCK6_KERNEL],[test x"$install_real_avx2_block6" = x"yes"])
if test x"${install_real_avx2_block6}" = x"yes" ; then
AC_DEFINE([WITH_REAL_AVX2_BLOCK6_KERNEL],[1],[can use real_avx2_block6 kernel])
fi
AM_CONDITIONAL([WITH_COMPLEX_SSE_BLOCK1_KERNEL],[test x"$install_complex_sse_block1" = x"yes"])
if test x"${install_complex_sse_block1}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_SSE_BLOCK1_KERNEL],[1],[can use complex_sse_block1 kernel])
fi
AM_CONDITIONAL([WITH_COMPLEX_SSE_BLOCK2_KERNEL],[test x"$install_complex_sse_block2" = x"yes"])
if test x"${install_complex_sse_block2}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_SSE_BLOCK2_KERNEL],[1],[can use complex_sse_block2 kernel])
fi
AM_CONDITIONAL([WITH_COMPLEX_AVX_BLOCK1_KERNEL],[test x"$install_complex_avx_block1" = x"yes"]) AM_CONDITIONAL([WITH_COMPLEX_AVX_BLOCK1_KERNEL],[test x"$install_complex_avx_block1" = x"yes"])
if test x"${install_complex_avx_block1}" = x"yes" ; then if test x"${install_complex_avx_block1}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_AVX_BLOCK1_KERNEL],[1],[can use complex_avx_block1 kernel]) AC_DEFINE([WITH_COMPLEX_AVX_BLOCK1_KERNEL],[1],[can use complex_avx_block1 kernel])
...@@ -736,6 +832,16 @@ if test x"${install_complex_avx_block2}" = x"yes" ; then ...@@ -736,6 +832,16 @@ if test x"${install_complex_avx_block2}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_AVX_BLOCK2_KERNEL],[1],[can use complex_avx_block2 kernel]) AC_DEFINE([WITH_COMPLEX_AVX_BLOCK2_KERNEL],[1],[can use complex_avx_block2 kernel])
fi fi
AM_CONDITIONAL([WITH_COMPLEX_AVX2_BLOCK1_KERNEL],[test x"$install_complex_avx2_block1" = x"yes"])
if test x"${install_complex_avx2_block1}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_AVX2_BLOCK1_KERNEL],[1],[can use complex_avx2_block1 kernel])
fi
AM_CONDITIONAL([WITH_COMPLEX_AVX2_BLOCK2_KERNEL],[test x"$install_complex_avx2_block2" = x"yes"])
if test x"${install_complex_avx2_block2}" = x"yes" ; then
AC_DEFINE([WITH_COMPLEX_AVX2_BLOCK2_KERNEL],[1],[can use complex_avx2_block2 kernel])
fi
AM_CONDITIONAL([WITH_REAL_BGP_KERNEL],[test x"$install_real_bgp" = x"yes"]) AM_CONDITIONAL([WITH_REAL_BGP_KERNEL],[test x"$install_real_bgp" = x"yes"])
if test x"${install_real_bgp}" = x"yes" ; then if test x"${install_real_bgp}" = x"yes" ; then
AC_DEFINE([WITH_REAL_BGP_KERNEL],[1],[can use real BGP kernel]) AC_DEFINE([WITH_REAL_BGP_KERNEL],[1],[can use real BGP kernel])
...@@ -840,14 +946,14 @@ mkdir -p test/shared_sources ...@@ -840,14 +946,14 @@ mkdir -p test/shared_sources
grep -h "^ *!c>" $srcdir/test/shared_sources/*.F90 | sed 's/^ *!c>//;' > test/shared_sources/generated.h || exit 1 grep -h "^ *!c>" $srcdir/test/shared_sources/*.F90 | sed 's/^ *!c>//;' > test/shared_sources/generated.h || exit 1
if test "${can_compile_avx}" = "no" ; then if test "${can_compile_avx}" = "no" ; then
if test x"${want_avx}" = x"yes" ; then # if test x"${want_avx}" = x"yes" ; then
AC_MSG_WARN([Could not compile AVX instructions]) AC_MSG_WARN([Could not compile AVX instructions])
fi # fi
fi fi
if test "${can_compile_avx2}" = "no" ; then if test "${can_compile_avx2}" = "no" ; then
if test x"${want_avx}" = x"yes" ; then # if test x"${want_avx2}" = x"yes" ; then
AC_MSG_WARN([Could not compile AVX2 instructions]) AC_MSG_WARN([Could not compile AVX2 instructions])
fi # fi
fi fi
if test "${can_compile_sse}" = "no" ; then if test "${can_compile_sse}" = "no" ; then
......
...@@ -3,21 +3,30 @@ ...@@ -3,21 +3,30 @@
#define ELPA2_REAL_KERNEL_BGP 3 #define ELPA2_REAL_KERNEL_BGP 3
#define ELPA2_REAL_KERNEL_BGQ 4 #define ELPA2_REAL_KERNEL_BGQ 4
#define ELPA2_REAL_KERNEL_SSE 5 #define ELPA2_REAL_KERNEL_SSE 5
#define ELPA2_REAL_KERNEL_AVX_BLOCK2 6 #define ELPA2_REAL_KERNEL_SSE_BLOCK2 6
#define ELPA2_REAL_KERNEL_AVX_BLOCK4 7 #define ELPA2_REAL_KERNEL_SSE_BLOCK4 7
#define ELPA2_REAL_KERNEL_AVX_BLOCK6 8 #define ELPA2_REAL_KERNEL_SSE_BLOCK6 8
#define ELPA2_REAL_KERNEL_GPU 9 #define ELPA2_REAL_KERNEL_AVX_BLOCK2 9
#define ELPA2_REAL_KERNEL_AVX_BLOCK4 10
#define ELPA2_NUMBER_OF_REAL_KERNELS 9 #define ELPA2_REAL_KERNEL_AVX_BLOCK6 11
#define ELPA2_REAL_KERNEL_AVX2_BLOCK2 12
#define ELPA2_REAL_KERNEL_AVX2_BLOCK4 13
#define ELPA2_REAL_KERNEL_AVX2_BLOCK6 14
#define ELPA2_REAL_KERNEL_GPU 15
#define ELPA2_NUMBER_OF_REAL_KERNELS 15
#define ELPA2_COMPLEX_KERNEL_GENERIC 1 #define ELPA2_COMPLEX_KERNEL_GENERIC 1
#define ELPA2_COMPLEX_KERNEL_GENERIC_SIMPLE 2 #define ELPA2_COMPLEX_KERNEL_GENERIC_SIMPLE 2
#define ELPA2_COMPLEX_KERNEL_BGP 3 #define ELPA2_COMPLEX_KERNEL_BGP 3
#define ELPA2_COMPLEX_KERNEL_BGQ 4 #define ELPA2_COMPLEX_KERNEL_BGQ 4
#define ELPA2_COMPLEX_KERNEL_SSE 5 #define ELPA2_COMPLEX_KERNEL_SSE 5
#define ELPA2_COMPLEX_KERNEL_AVX_BLOCK1 6 #define ELPA2_COMPLEX_KERNEL_SSE_BLOCK1 6
#define ELPA2_COMPLEX_KERNEL_AVX_BLOCK2 7 #define ELPA2_COMPLEX_KERNEL_SSE_BLOCK2 7
#define ELPA2_COMPLEX_KERNEL_GPU 8 #define ELPA2_COMPLEX_KERNEL_AVX_BLOCK1 8
#define ELPA2_COMPLEX_KERNEL_AVX_BLOCK2 9
#define ELPA2_COMPLEX_KERNEL_AVX2_BLOCK1 10
#define ELPA2_COMPLEX_KERNEL_AVX2_BLOCK2 11
#define ELPA2_COMPLEX_KERNEL_GPU 12
#define ELPA2_NUMBER_OF_COMPLEX_KERNELS 8 #define ELPA2_NUMBER_OF_COMPLEX_KERNELS 12
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
...@@ -73,14 +73,21 @@ module ELPA2_utilities ...@@ -73,14 +73,21 @@ module ELPA2_utilities
public :: get_actual_real_kernel_name, get_actual_complex_kernel_name public :: get_actual_real_kernel_name, get_actual_complex_kernel_name
public :: REAL_ELPA_KERNEL_GENERIC, REAL_ELPA_KERNEL_GENERIC_SIMPLE, & public :: REAL_ELPA_KERNEL_GENERIC, REAL_ELPA_KERNEL_GENERIC_SIMPLE, &
REAL_ELPA_KERNEL_BGP, REAL_ELPA_KERNEL_BGQ, & REAL_ELPA_KERNEL_BGP, REAL_ELPA_KERNEL_BGQ, &
REAL_ELPA_KERNEL_SSE, REAL_ELPA_KERNEL_AVX_BLOCK2, & REAL_ELPA_KERNEL_SSE, REAL_ELPA_KERNEL_SSE_BLOCK2, &
REAL_ELPA_KERNEL_SSE_BLOCK4, REAL_ELPA_KERNEL_SSE_BLOCK6, &
REAL_ELPA_KERNEL_AVX_BLOCK2, &
REAL_ELPA_KERNEL_AVX_BLOCK4, REAL_ELPA_KERNEL_AVX_BLOCK6, & REAL_ELPA_KERNEL_AVX_BLOCK4, REAL_ELPA_KERNEL_AVX_BLOCK6, &
REAL_ELPA_KERNEL_AVX2_BLOCK2, &
REAL_ELPA_KERNEL_AVX2_BLOCK4, REAL_ELPA_KERNEL_AVX2_BLOCK6, &
REAL_ELPA_KERNEL_GPU REAL_ELPA_KERNEL_GPU
public :: COMPLEX_ELPA_KERNEL_GENERIC, COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE, & public :: COMPLEX_ELPA_KERNEL_GENERIC, COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE, &
COMPLEX_ELPA_KERNEL_BGP, COMPLEX_ELPA_KERNEL_BGQ, & COMPLEX_ELPA_KERNEL_BGP, COMPLEX_ELPA_KERNEL_BGQ, &
COMPLEX_ELPA_KERNEL_SSE, COMPLEX_ELPA_KERNEL_AVX_BLOCK1, & COMPLEX_ELPA_KERNEL_SSE, COMPLEX_ELPA_KERNEL_SSE_BLOCK1, &
COMPLEX_ELPA_KERNEL_AVX_BLOCK2, & COMPLEX_ELPA_KERNEL_SSE_BLOCK2, &
COMPLEX_ELPA_KERNEL_AVX_BLOCK1,COMPLEX_ELPA_KERNEL_AVX_BLOCK2, &
COMPLEX_ELPA_KERNEL_AVX2_BLOCK1,COMPLEX_ELPA_KERNEL_AVX2_BLOCK2, &
COMPLEX_ELPA_KERNEL_GPU COMPLEX_ELPA_KERNEL_GPU
public :: REAL_ELPA_KERNEL_NAMES, COMPLEX_ELPA_KERNEL_NAMES public :: REAL_ELPA_KERNEL_NAMES, COMPLEX_ELPA_KERNEL_NAMES
...@@ -95,17 +102,22 @@ module ELPA2_utilities ...@@ -95,17 +102,22 @@ module ELPA2_utilities
public :: qr_decomposition_via_environment_variable public :: qr_decomposition_via_environment_variable
integer(kind=ik), parameter :: number_of_real_kernels = ELPA2_NUMBER_OF_REAL_KERNELS integer, parameter :: number_of_real_kernels = ELPA2_NUMBER_OF_REAL_KERNELS
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC = ELPA2_REAL_KERNEL_GENERIC integer, parameter :: REAL_ELPA_KERNEL_GENERIC = ELPA2_REAL_KERNEL_GENERIC
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GENERIC_SIMPLE = ELPA2_REAL_KERNEL_GENERIC_SIMPLE integer, parameter :: REAL_ELPA_KERNEL_GENERIC_SIMPLE = ELPA2_REAL_KERNEL_GENERIC_SIMPLE
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGP = ELPA2_REAL_KERNEL_BGP integer, parameter :: REAL_ELPA_KERNEL_BGP = ELPA2_REAL_KERNEL_BGP
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_BGQ = ELPA2_REAL_KERNEL_BGQ integer, parameter :: REAL_ELPA_KERNEL_BGQ = ELPA2_REAL_KERNEL_BGQ
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_SSE = ELPA2_REAL_KERNEL_SSE integer, parameter :: REAL_ELPA_KERNEL_SSE = ELPA2_REAL_KERNEL_SSE
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_AVX_BLOCK2 = ELPA2_REAL_KERNEL_AVX_BLOCK2 integer, parameter :: REAL_ELPA_KERNEL_SSE_BLOCK2 = ELPA2_REAL_KERNEL_SSE_BLOCK2
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_AVX_BLOCK4 = ELPA2_REAL_KERNEL_AVX_BLOCK4 integer, parameter :: REAL_ELPA_KERNEL_SSE_BLOCK4 = ELPA2_REAL_KERNEL_SSE_BLOCK4
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_AVX_BLOCK6 = ELPA2_REAL_KERNEL_AVX_BLOCK6 integer, parameter :: REAL_ELPA_KERNEL_SSE_BLOCK6 = ELPA2_REAL_KERNEL_SSE_BLOCK6
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GPU = ELPA2_REAL_KERNEL_GPU integer, parameter :: REAL_ELPA_KERNEL_AVX_BLOCK2 = ELPA2_REAL_KERNEL_AVX_BLOCK2
integer, parameter :: REAL_ELPA_KERNEL_AVX_BLOCK4 = ELPA2_REAL_KERNEL_AVX_BLOCK4
integer, parameter :: REAL_ELPA_KERNEL_AVX_BLOCK6 = ELPA2_REAL_KERNEL_AVX_BLOCK6
integer, parameter :: REAL_ELPA_KERNEL_AVX2_BLOCK2 = ELPA2_REAL_KERNEL_AVX2_BLOCK2
integer, parameter :: REAL_ELPA_KERNEL_AVX2_BLOCK4 = ELPA2_REAL_KERNEL_AVX2_BLOCK4
integer, parameter :: REAL_ELPA_KERNEL_AVX2_BLOCK6 = ELPA2_REAL_KERNEL_AVX2_BLOCK6
integer(kind=ik), parameter :: REAL_ELPA_KERNEL_GPU = ELPA2_REAL_KERNEL_GPU
#if defined(WITH_REAL_AVX_BLOCK2_KERNEL) #if defined(WITH_REAL_AVX_BLOCK2_KERNEL)
...@@ -187,21 +199,30 @@ module ELPA2_utilities ...@@ -187,21 +199,30 @@ module ELPA2_utilities
"REAL_ELPA_KERNEL_BGP ", & "REAL_ELPA_KERNEL_BGP ", &
"REAL_ELPA_KERNEL_BGQ ", & "REAL_ELPA_KERNEL_BGQ ", &
"REAL_ELPA_KERNEL_SSE ", & "REAL_ELPA_KERNEL_SSE ", &
"REAL_ELPA_KERNEL_SSE_BLOCK2 ", &
"REAL_ELPA_KERNEL_SSE_BLOCK4 ", &
"REAL_ELPA_KERNEL_SSE_BLOCK6 ", &
"REAL_ELPA_KERNEL_AVX_BLOCK2 ", & "REAL_ELPA_KERNEL_AVX_BLOCK2 ", &
"REAL_ELPA_KERNEL_AVX_BLOCK4 ", & "REAL_ELPA_KERNEL_AVX_BLOCK4 ", &
"REAL_ELPA_KERNEL_AVX_BLOCK6 ", & "REAL_ELPA_KERNEL_AVX_BLOCK6 ", &
"REAL_ELPA_KERNEL_AVX2_BLOCK2 ", &
"REAL_ELPA_KERNEL_AVX2_BLOCK4 ", &
"REAL_ELPA_KERNEL_AVX2_BLOCK6 ", &
"REAL_ELPA_KERNEL_GPU "/) "REAL_ELPA_KERNEL_GPU "/)
integer(kind=ik), parameter :: number_of_complex_kernels = ELPA2_NUMBER_OF_COMPLEX_KERNELS integer, parameter :: number_of_complex_kernels = ELPA2_NUMBER_OF_COMPLEX_KERNELS
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_GENERIC = ELPA2_COMPLEX_KERNEL_GENERIC integer, parameter :: COMPLEX_ELPA_KERNEL_GENERIC = ELPA2_COMPLEX_KERNEL_GENERIC
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE = ELPA2_COMPLEX_KERNEL_GENERIC_SIMPLE integer, parameter :: COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE = ELPA2_COMPLEX_KERNEL_GENERIC_SIMPLE
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_BGP = ELPA2_COMPLEX_KERNEL_BGP integer, parameter :: COMPLEX_ELPA_KERNEL_BGP = ELPA2_COMPLEX_KERNEL_BGP
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_BGQ = ELPA2_COMPLEX_KERNEL_BGQ integer, parameter :: COMPLEX_ELPA_KERNEL_BGQ = ELPA2_COMPLEX_KERNEL_BGQ
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_SSE = ELPA2_COMPLEX_KERNEL_SSE integer, parameter :: COMPLEX_ELPA_KERNEL_SSE = ELPA2_COMPLEX_KERNEL_SSE
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_AVX_BLOCK1 = ELPA2_COMPLEX_KERNEL_AVX_BLOCK1 integer, parameter :: COMPLEX_ELPA_KERNEL_SSE_BLOCK1 = ELPA2_COMPLEX_KERNEL_SSE_BLOCK1
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_AVX_BLOCK2 = ELPA2_COMPLEX_KERNEL_AVX_BLOCK2 integer, parameter :: COMPLEX_ELPA_KERNEL_SSE_BLOCK2 = ELPA2_COMPLEX_KERNEL_SSE_BLOCK2
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_GPU = ELPA2_COMPLEX_KERNEL_GPU integer, parameter :: COMPLEX_ELPA_KERNEL_AVX_BLOCK1 = ELPA2_COMPLEX_KERNEL_AVX_BLOCK1
integer, parameter :: COMPLEX_ELPA_KERNEL_AVX_BLOCK2 = ELPA2_COMPLEX_KERNEL_AVX_BLOCK2
integer, parameter :: COMPLEX_ELPA_KERNEL_AVX2_BLOCK1 = ELPA2_COMPLEX_KERNEL_AVX2_BLOCK1
integer, parameter :: COMPLEX_ELPA_KERNEL_AVX2_BLOCK2 = ELPA2_COMPLEX_KERNEL_AVX2_BLOCK2
integer(kind=ik), parameter :: COMPLEX_ELPA_KERNEL_GPU = ELPA2_COMPLEX_KERNEL_GPU
#if defined(WITH_COMPLEX_AVX_BLOCK1_KERNEL) #if defined(WITH_COMPLEX_AVX_BLOCK1_KERNEL)
...@@ -268,8 +289,12 @@ module ELPA2_utilities ...@@ -268,8 +289,12 @@ module ELPA2_utilities
"COMPLEX_ELPA_KERNEL_BGP ", & "COMPLEX_ELPA_KERNEL_BGP ", &
"COMPLEX_ELPA_KERNEL_BGQ ", & "COMPLEX_ELPA_KERNEL_BGQ ", &
"COMPLEX_ELPA_KERNEL_SSE ", & "COMPLEX_ELPA_KERNEL_SSE ", &
"COMPLEX_ELPA_KERNEL_SSE_BLOCK1 ", &
"COMPLEX_ELPA_KERNEL_SSE_BLOCK2 ", &
"COMPLEX_ELPA_KERNEL_AVX_BLOCK1 ", & "COMPLEX_ELPA_KERNEL_AVX_BLOCK1 ", &
"COMPLEX_ELPA_KERNEL_AVX_BLOCK2 ", & "COMPLEX_ELPA_KERNEL_AVX_BLOCK2 ", &
"COMPLEX_ELPA_KERNEL_AVX2_BLOCK1 ", &
"COMPLEX_ELPA_KERNEL_AVX2_BLOCK2 ", &
"COMPLEX_ELPA_KERNEL_GPU "/) "COMPLEX_ELPA_KERNEL_GPU "/)
integer(kind=ik), parameter :: & integer(kind=ik), parameter :: &
...@@ -300,27 +325,58 @@ module ELPA2_utilities ...@@ -300,27 +325,58 @@ module ELPA2_utilities
#else #else
,0 & ,0 &
#endif #endif
#if WITH_REAL_AVX_BLOCK2_KERNEL #if WITH_REAL_SSE_BLOCK2_KERNEL
,1 & ,1 &
#else #else
,0 & ,0 &
#endif #endif
#if WITH_REAL_AVX_BLOCK4_KERNEL #if WITH_REAL_SSE_BLOCK4_KERNEL
,1 & ,1 &
#else #else
,0 & ,0 &
#endif #endif
#if WITH_REAL_AVX_BLOCK6_KERNEL #if WITH_REAL_SSE_BLOCK6_KERNEL
,1 & ,1 &
#else #else
,0 & ,0 &
#endif #endif
#ifdef WITH_GPU_VERSION #if WITH_REAL_AVX_BLOCK2_KERNEL
,1 & ,1 &
#else #else
,0 & ,0 &
#endif #endif
/) #if WITH_REAL_AVX_BLOCK4_KERNEL
,1 &
#else
,0 &
#endif
#if WITH_REAL_AVX_BLOCK6_KERNEL
,1 &
#else
,0 &
#endif
#if WITH_REAL_AVX2_BLOCK2_KERNEL
,1 &
#else
,0 &
#endif
#if WITH_REAL_AVX2_BLOCK4_KERNEL
,1 &
#else
,0 &
#endif
#if WITH_REAL_AVX2_BLOCK6_KERNEL
,1 &
#else
,0 &
#endif
#ifdef WITH_GPU_VERSION
,1 &
#else
,0 &