Merge branch 'ELPA_GPU' into ELPA_KNL

.gitlab-ci.yml

@@ -710,7 +710,7 @@ gfortran-single-precision-nompi-noopenmp-ftimings-assumed-size-jobs:
     - cpu
   script:
     - ./autogen.sh
-    - ./configure FC=gfortran CFLAGS="--coverage -O3 -mavx" CXXFLAGS="--coverage -O3 -mavx" FCFLAGS="--coverage -O3 -mavx" SCALAPACK_LDFLAGS="$MKL_GFORTRAN_SCALAPACK_LDFLAGS_NO_MPI_NO_OMP" SCALAPACK_FCFLAGS="$MKL_GFORTRAN_SCALAPACK_FCFLAGS_NO_MPI_NO_OMP" --with-mpi=0 --enable-single-precision --with-ftimings --emable-assumed-size-arrays
+    - ./configure FC=gfortran CFLAGS="--coverage -O3 -mavx" CXXFLAGS="--coverage -O3 -mavx" FCFLAGS="--coverage -O3 -mavx" SCALAPACK_LDFLAGS="$MKL_GFORTRAN_SCALAPACK_LDFLAGS_NO_MPI_NO_OMP" SCALAPACK_FCFLAGS="$MKL_GFORTRAN_SCALAPACK_FCFLAGS_NO_MPI_NO_OMP" --with-mpi=0 --enable-single-precision --with-ftimings --enable-assumed-size-arrays
     - make -j 8
     - export LD_LIBRARY_PATH=$MKL_HOME/lib/intel64:$LD_LIBRARY_PATH
     - make check TEST_FLAGS='1500 50 16'

src/elpa2_kernels/elpa2_kernels_complex_avx-avx2_1hv_single_precision.c

@@ -73,7 +73,7 @@ static  __forceinline void hh_trafo_complex_kernel_8_AVX_1hv_single(float comple
 static  __forceinline void hh_trafo_complex_kernel_4_AVX_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
 
 /*
-!f>#ifdef HAVE_AVX
+!f>#if defined(HAVE_AVX) || defined(HAVE_AVX2)
 !f> interface
 !f>   subroutine single_hh_trafo_complex_avx_avx2_1hv_single(q, hh, pnb, pnq, pldq) &
 !f>                             bind(C, name="single_hh_trafo_complex_avx_avx2_1hv_single")

src/elpa2_kernels/elpa2_kernels_complex_avx-avx2_2hv_single_precision.c

@@ -74,7 +74,7 @@ static __forceinline void hh_trafo_complex_kernel_4_AVX_2hv_single(float complex
 //static __forceinline void hh_trafo_complex_kernel_2_AVX_2hv_single(float complex* q, float complex* hh, int nb, int ldq, int ldh, float complex s, float complex s1);
 
 /*
-!f>#ifdef HAVE_AVX
+!f>#if defined(HAVE_AVX) || defined(HAVE_AVX2)
 !f> interface
 !f>   subroutine double_hh_trafo_complex_avx_avx2_2hv_single(q, hh, pnb, pnq, pldq, pldh) &
 !f>                             bind(C, name="double_hh_trafo_complex_avx_avx2_2hv_single")

src/elpa2_kernels/elpa2_kernels_real_avx-avx2_2hv_single_precision.c

@@ -74,7 +74,7 @@ __forceinline void hh_trafo_kernel_24_AVX_2hv_single(float* q, float* hh, int nb
 void double_hh_trafo_real_avx_avx2_2hv_single(float* q, float* hh, int* pnb, int* pnq, int* pldq, int* pldh);
 
 /*
-!f>#ifdef HAVE_AVX
+!f>#if defined(HAVE_AVX) || defined(HAVE_AVX2)
 !f> interface
 !f>   subroutine double_hh_trafo_real_avx_avx2_2hv_single(q, hh, pnb, pnq, pldq, pldh) &
 !f>                             bind(C, name="double_hh_trafo_real_avx_avx2_2hv_single")

src/elpa2_kernels/elpa2_kernels_real_avx-avx2_4hv_single_precision.c

@@ -78,7 +78,7 @@ __forceinline void hh_trafo_kernel_24_AVX_4hv_single(float* q, float* hh, int nb
 void quad_hh_trafo_real_avx_avx2_4hv_single(float* q, float* hh, int* pnb, int* pnq, int* pldq, int* pldh);
 
 /*
-!f>#ifdef HAVE_AVX
+!f>#if defined(HAVE_AVX) || defined(HAVE_AVX2)
 !f> interface
 !f>   subroutine quad_hh_trafo_real_avx_avx2_4hv_single(q, hh, pnb, pnq, pldq, pldh) &
 !f>                             bind(C, name="quad_hh_trafo_real_avx_avx2_4hv_single")

src/elpa2_kernels/elpa2_kernels_real_avx-avx2_6hv_single_precision.c

@@ -75,7 +75,7 @@ static void hh_trafo_kernel_8_AVX_6hv_single(float* q, float* hh, int nb, int ld
 void hexa_hh_trafo_real_avx_avx2_6hv_single_(float* q, float* hh, int* pnb, int* pnq, int* pldq, int* pldh);
 
 /*
-!f>#ifdef HAVE_AVX
+!f>#if defined(HAVE_AVX) || defined(HAVE_AVX2)
 !f> interface
 !f>   subroutine hexa_hh_trafo_real_avx_avx2_6hv_single(q, hh, pnb, pnq, pldq, pldh) &
 !f>                             bind(C, name="hexa_hh_trafo_real_avx_avx2_6hv_single")