Rename test programs and CI tests

Makefile.am

@@ -784,6 +784,7 @@ CLEANFILES = \
 clean-local:
 	-rm -rf modules/* private_modules/* test_modules/* .fortran_dependencies/*
 	-rm -rf validate_*.sh
+	-rm -rf test_*.sh
 	-rm -rf real_2stage*.sh
 	-rm -rf complex_2stage*.sh
 	-rm -rf single_complex_2stage*.sh

ci_test_scripts/generate_gitlab_ci_tests.py

@@ -346,7 +346,7 @@ ilp64_no_omp_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_NOMPI_NOOMP_ILP64 \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_FCFLAGS_NOMPI_NOOMP_ILP64 \\" '
     '--enable-option-checking=fatal --with-mpi=no --disable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -364,7 +364,7 @@ ilp64_no_omp_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_NOMPI_OMP_ILP64 \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_FCFLAGS_NOMPI_OMP_ILP64 \\" '
     '--enable-option-checking=fatal --with-mpi=no --enable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -388,7 +388,7 @@ ilp64_no_omp_mpi_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_MPI_NOOMP_ILP64 \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_FCFLAGS_MPI_NOOMP_ILP64 \\" '
     '--enable-option-checking=fatal --with-mpi=yes --disable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -406,7 +406,7 @@ ilp64_no_omp_mpi_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_MPI_OMP_ILP64 \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_FCFLAGS_MPI_OMP_ILP64 \\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-64bit-integer-math-support || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -430,7 +430,7 @@ matrix_redistribute_mpi_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_MPI_NO_OMP \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_MPI_NO_OMP \\" '
     '--enable-option-checking=fatal --with-mpi=yes --disable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-scalapack-tests --enable-autotune-redistribute-matrix || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-scalapack-tests --enable-autotune-redistribute-matrix || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -448,7 +448,7 @@ matrix_redistribute_mpi_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_LDFLAGS_MPI_OMP \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_GFORTRAN_SCALAPACK_FCFLAGS_MPI_OMP \\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-scalapack-tests --enable-autotune-redistribute-matrix || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --disable-avx2 --disable-avx512 --enable-scalapack-tests --enable-autotune-redistribute-matrix || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -472,7 +472,7 @@ python_ci_tests = [
     'SCALAPACK_LDFLAGS=\\"$MKL_ANACONDA_INTEL_SCALAPACK_LDFLAGS_MPI_OMP \\" '
     'SCALAPACK_FCFLAGS=\\"$MKL_ANACONDA_INTEL_SCALAPACK_FCFLAGS_MPI_OMP \\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --enable-python --enable-python-tests || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --enable-python --enable-python-tests || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m $MATRIX_SIZE -n $NUMBER_OF_EIGENVECTORS -b $BLOCK_SIZE '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -486,7 +486,7 @@ python_ci_tests = [
     'SCALAPACK_LDFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_LDFLAGS_MPI_OMP\\\" '
     'SCALAPACK_FCFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_FCFLAGS_MPI_OMP\\\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --enable-python --enable-python-tests || { cat config.log; exit 1; }'
+    '--disable-nvidia-gpu --enable-avx --enable-python --enable-python-tests || { cat config.log; exit 1; }'
     '" -j 8 -t $MPI_TASKS -m 150 -n 50 -b 16 '
     '-s $SKIP_STEP -i $INTERACTIVE_RUN -S $SLURM',
     "\n",
@@ -496,13 +496,13 @@ python_ci_tests = [
     'SCALAPACK_LDFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_LDFLAGS_MPI_OMP \\\" '
     'SCALAPACK_FCFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_FCFLAGS_MPI_OMP \\\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --enable-python --enable-python-tests "'
+    '--disable-nvidia-gpu --enable-avx --enable-python --enable-python-tests "'
     '-d " CC=\\\\\\\"mpiicc\\\\\\\" CFLAGS=\\\\\\\"-O3 -xAVX\\\\\\\" '
     'FC=\\\\\\\"mpiifort\\\\\\\" FCFLAGS=\\\\\\\"-O3 -xAVX\\\\\\\" '
     'SCALAPACK_LDFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_LDFLAGS_MPI_OMP \\\" '
     'SCALAPACK_FCFLAGS=\\\"$MKL_ANACONDA_INTEL_SCALAPACK_FCFLAGS_MPI_OMP \\\" '
     '--enable-option-checking=fatal --with-mpi=yes --enable-openmp '
-    '--disable-gpu --enable-avx --enable-python --enable-python-tests'
+    '--disable-nvidia-gpu --enable-avx --enable-python --enable-python-tests'
     '" -t $MPI_TASKS -m 150 -n 50 -b 16 -S $SLURM  || { chmod u+rwX -R . ; exit 1 ; }',
     "\n",
 ]
@@ -597,8 +597,8 @@ band_to_full_blocking = {
 }
 
 gpu = {
-        "no-gpu" : "--disable-gpu",
-        "with-gpu" : "--enable-gpu --with-cuda-path=\\$CUDA_HOME/",
+        "no-gpu" : "--disable-nvidia-gpu",
+        "with-gpu" : "--enable-nvidia-gpu --with-cuda-path=\\$CUDA_HOME/",
 }
 
 

generate_automake_test_programs.py

@@ -21,8 +21,9 @@ solver_flag = {
     "scalapack_part": "-DTEST_SCALAPACK_PART",
 }
 gpu_flag = {
-    0: "-DTEST_GPU=0",
-    1: "-DTEST_GPU=1",
+    "GPU_OFF": "-DTEST_NVIDIA_GPU=0 -DTEST_AMD_GPU=0",
+    "NVIDIA_GPU_ON": "-DTEST_NVIDIA_GPU=1",
+    "AMD_GPU_ON": "-DTEST_AMD_GPU=1",
 }
 gpu_id_flag = {
     0: "-DTEST_GPU_SET_ID=0",
@@ -86,14 +87,14 @@ for lang, m, g, gid, q, t, p, d, s, lay, spl in product(sorted(language_flag.key
     # exclude some test combinations
 
     # analytic tests only for "eigenvectors" and not on GPU
-    if(m == "analytic" and (g == 1 or t != "eigenvectors")):
+    if(m == "analytic" and ( g == "NVIDIA_GPU_ON" or g == "AMD_GPU_ON" or t != "eigenvectors")):
         continue
 
     # Frank tests only for "eigenvectors" and eigenvalues and real double precision case
     if(m == "frank" and ((t != "eigenvectors" or t != "eigenvalues") and (d != "real" or p != "double"))):
         continue
 
-    if(s in ["scalapack_all", "scalapack_part"] and (g == 1 or t != "eigenvectors" or m != "analytic")):
+    if(s in ["scalapack_all", "scalapack_part"] and (g == "NVIDIA_GPU_ON" or g == "AMD_GPU_ON" or t != "eigenvectors" or m != "analytic")):
         continue
 
     # do not test single-precision scalapack
@@ -127,7 +128,7 @@ for lang, m, g, gid, q, t, p, d, s, lay, spl in product(sorted(language_flag.key
         continue
 
     # qr only for 2stage real
-    if (q == 1 and (s != "2stage" or d != "real" or t != "eigenvectors" or g == 1 or m != "random")):
+    if (q == 1 and (s != "2stage" or d != "real" or t != "eigenvectors" or g == "NVIDIA_GPU_ON" or "AMD_GPU_ON" or m != "random")):
         continue
 
     if(spl == "myself" and (d != "real" or p != "double" or q != 0 or m != "random" or (t != "eigenvectors" and t != "cholesky")  or lang != "Fortran" or lay != "square")):
@@ -154,10 +155,14 @@ for lang, m, g, gid, q, t, p, d, s, lay, spl in product(sorted(language_flag.key
             print("if ENABLE_C_TESTS")
             endifs += 1
 
-        if (g == 1):
+        if (g == "NVIDIA_GPU_ON"):
             print("if WITH_NVIDIA_GPU_VERSION")
             endifs += 1
 
+        if (g == "AMD_GPU_ON"):
+            print("if WITH_AMD_GPU_VERSION")
+            endifs += 1
+
         if (lay == "all_layouts"):
             print("if WITH_MPI")
             endifs += 1
@@ -190,7 +195,7 @@ for lang, m, g, gid, q, t, p, d, s, lay, spl in product(sorted(language_flag.key
             langsuffix=language_flag[lang],
             d=d, p=p, t=t, s=s,
             kernelsuffix="" if kernel == "nokernel" else "_" + kernel,
-            gpusuffix="gpu_" if g else "",
+            gpusuffix="gpu_" if  (g == "NVIDIA_GPU_ON" or g == "AMD_GPU_ON") else "",
             gpuidsuffix="set_gpu_id_" if gid else "",
             qrsuffix="qr_" if q else "",
             m=m,

test/C/test.c

@@ -127,6 +127,14 @@
 #define TEST_C_INT_MPI_TYPE int
 #define C_INT_MPI_TYPE int
 #endif
+
+#define TEST_GPU  0
+#if (TEST_NVIDIA_GPU == 1) || (TEST_AMD_GPU == 1)
+#undef TEST_GPU
+#define TEST_GPU  1
+#endif
+
+
 #include "test/shared/generated.h"
 
 int main(int argc, char** argv) {
@@ -273,8 +281,17 @@ int main(int argc, char** argv) {
 #endif
    assert_elpa_ok(error_elpa);
 
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)   
    elpa_set(handle, "nvidia-gpu", TEST_GPU, &error_elpa);
    assert_elpa_ok(error_elpa);
+#endif
+
+#if TEST_AMD_GPU == 1
+   elpa_set(handle, "amd-gpu", TEST_GPU, &error_elpa);
+   assert_elpa_ok(error_elpa);
+#endif
+
+
 
 #if defined(TEST_SOLVE_2STAGE) && defined(TEST_KERNEL)
 # ifdef TEST_COMPLEX

test/Fortran/test.F90

@@ -45,7 +45,8 @@
 ! Define one of TEST_REAL or TEST_COMPLEX
 ! Define one of TEST_SINGLE or TEST_DOUBLE
 ! Define one of TEST_SOLVER_1STAGE or TEST_SOLVER_2STAGE
-! Define TEST_GPU \in [0, 1]
+! Define TEST_NVIDIA_GPU \in [0, 1]
+! Define TEST_AMD_GPU \in [0, 1]
 ! Define either TEST_ALL_KERNELS or a TEST_KERNEL \in [any valid kernel]
 
 #if !(defined(TEST_REAL) ^ defined(TEST_COMPLEX))
@@ -116,6 +117,13 @@ error: define either TEST_ALL_KERNELS or a valid TEST_KERNEL
 #define TEST_INT_MPI_TYPE integer(kind=c_int32_t)
 #define INT_MPI_TYPE c_int32_t
 #endif
+
+#define TEST_GPU 0
+#if (TEST_NVIDIA_GPU == 1) || (TEST_AMD_GPU == 1)
+#undef TEST_GPU
+#define TEST_GPU 1
+#endif
+
 #include "assert.h"
 
 program test
@@ -640,8 +648,15 @@ program test
 #endif
    assert_elpa_ok(error_elpa)
 
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)
    call e%set("nvidia-gpu", TEST_GPU, error_elpa)
    assert_elpa_ok(error_elpa)
+#endif
+
+#if TEST_AMD_GPU == 1
+   call e%set("amd-gpu", TEST_GPU, error_elpa)
+   assert_elpa_ok(error_elpa)
+#endif
 
 #if TEST_GPU_SET_ID == 1
    ! simple test
@@ -681,12 +696,24 @@ program test
 #ifdef TEST_SOLVER_2STAGE
 #if TEST_GPU == 1
 #if defined TEST_REAL
+#if (TEST_NVIDIA_GPU == 1)
      kernel = ELPA_2STAGE_REAL_NVIDIA_GPU
 #endif
+#if (TEST_AMD_GPU == 1)
+     kernel = ELPA_2STAGE_REAL_AND_GPU
+#endif
+#endif /* TEST_REAL */
 #if defined TEST_COMPLEX
+#if (TEST_NVIDIA_GPU == 1)
      kernel = ELPA_2STAGE_COMPLEX_NVIDIA_GPU
 #endif
+#if (TEST_AMD_GPU == 1)
+     kernel = ELPA_2STAGE_COMPLEX_AMD_GPU
 #endif
+#endif /* TEST_COMPLEX */
+#endif /* TEST_GPU == 1 */
+
+
      call e%set(KERNEL_KEY, kernel, error_elpa)
 #ifdef TEST_KERNEL
      assert_elpa_ok(error_elpa)

test/Fortran/test_autotune.F90

@@ -45,7 +45,8 @@
 ! Define one of TEST_REAL or TEST_COMPLEX
 ! Define one of TEST_SINGLE or TEST_DOUBLE
 ! Define one of TEST_SOLVER_1STAGE or TEST_SOLVER_2STAGE
-! Define TEST_GPU \in [0, 1]
+! Define TEST_NVIDIA_GPU \in [0, 1]
+! Define TEST_AMD_GPU \in [0, 1]
 ! Define either TEST_ALL_KERNELS or a TEST_KERNEL \in [any valid kernel]
 
 #if !(defined(TEST_REAL) ^ defined(TEST_COMPLEX))
@@ -94,6 +95,16 @@ error: define exactly one of TEST_SINGLE or TEST_DOUBLE
 #define TEST_INT_MPI_TYPE integer(kind=c_int32_t)
 #define INT_MPI_TYPE c_int32_t
 #endif
+
+
+
+#define TEST_GPU 0
+#if (TEST_NVIDIA_GPU == 1) || (TEST_AMD_GPU == 1)
+#undef TEST_GPU
+#define TEST_GPU 1
+#endif
+
+
 #include "assert.h"
 
 program test
@@ -228,8 +239,15 @@ program test
 
    call e%set("debug",1, error_elpa)
    assert_elpa_ok(error_elpa)
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)
    call e%set("nvidia-gpu", 0, error_elpa)
    assert_elpa_ok(error_elpa)
+#endif
+#if TEST_AMD_GPU == 1
+   call e%set("amd-gpu", 0, error_elpa)
+   assert_elpa_ok(error_elpa)
+#endif
+
    !call e%set("max_stored_rows", 15, error_elpa)
 
    assert_elpa_ok(e%setup())

test/Fortran/test_multiple_objs.F90

@@ -45,7 +45,8 @@
 ! Define one of TEST_REAL or TEST_COMPLEX
 ! Define one of TEST_SINGLE or TEST_DOUBLE
 ! Define one of TEST_SOLVER_1STAGE or TEST_SOLVER_2STAGE
-! Define TEST_GPU \in [0, 1]
+! Define TEST_NVIDIA_GPU \in [0, 1]
+! Define TEST_AMD_GPU \in [0, 1]
 ! Define either TEST_ALL_KERNELS or a TEST_KERNEL \in [any valid kernel]
 
 #if !(defined(TEST_REAL) ^ defined(TEST_COMPLEX))
@@ -95,6 +96,11 @@ error: define exactly one of TEST_SINGLE or TEST_DOUBLE
 #define INT_MPI_TYPE c_int32_t
 #endif
 
+#define TEST_GPU 0
+#if (TEST_NVIDIA_GPU == 1) || (TEST_AMD_GPU == 1)
+#undef TEST_GPU
+#define TEST_GPU 1
+#endif
 
 #include "assert.h"
 
@@ -209,8 +215,15 @@ program test
    call e1%set("debug",1, error_elpa)
    assert_elpa_ok(error_elpa)
 
-   call e1%set("nvidia-gpu", 0, error_elpa)
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)
+   call e1%set("nvidia-gpu", TEST_GPU, error_elpa)
+   assert_elpa_ok(error_elpa)
+#endif
+
+#if TEST_AMD_GPU == 1
+   call e1%set("amd-gpu", TEST_GPU, error_elpa)
    assert_elpa_ok(error_elpa)
+#endif
    !call e1%set("max_stored_rows", 15, error_elpa)
 
    assert_elpa_ok(e1%setup())
@@ -238,8 +251,14 @@ program test
    assert_elpa_ok(error_elpa)
    call e2%get("debug", int(debug,kind=c_int), error_elpa)
    assert_elpa_ok(error_elpa)
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)
    call e2%get("nvidia-gpu", int(gpu,kind=c_int), error_elpa)
    assert_elpa_ok(error_elpa)
+#endif
+#if TEST_NVIDIA_AMD == 1
+   call e2%get("amd-gpu", int(gpu,kind=c_int), error_elpa)
+   assert_elpa_ok(error_elpa)
+#endif
 
    if ((timings .ne. 1) .or. (debug .ne. 1) .or. (gpu .ne. 0)) then
      print *, "Parameters not stored or loaded correctly. Aborting...", timings, debug, gpu

test/Fortran/test_skewsymmetric.F90

@@ -45,7 +45,8 @@
 ! Define one of TEST_REAL or TEST_COMPLEX
 ! Define one of TEST_SINGLE or TEST_DOUBLE
 ! Define one of TEST_SOLVER_1STAGE or TEST_SOLVER_2STAGE
-! Define TEST_GPU \in [0, 1]
+! Define TEST_NVIDIA_GPU \in [0, 1]
+! Define TEST_AMD_GPU \in [0, 1]
 ! Define either TEST_ALL_KERNELS or a TEST_KERNEL \in [any valid kernel]
 
 #if !(defined(TEST_REAL) ^ defined(TEST_COMPLEX))
@@ -95,6 +96,13 @@ error: define exactly one of TEST_SINGLE or TEST_DOUBLE
 #define TEST_INT_MPI_TYPE integer(kind=c_int32_t)
 #define INT_MPI_TYPE c_int32_t
 #endif
+
+#define TEST_GPU 0
+#if (TEST_NVIDIA_GPU == 1) || (TEST_AMD_GPU == 1)
+#undef TEST_GPU
+#define TEST_GPU 1
+#endif
+
 #include "assert.h"
 
 program test
@@ -233,7 +241,13 @@ program test
    call e_complex%set("timings",1, error_elpa)
 
    call e_complex%set("debug",1,error_elpa)
-   call e_complex%set("nvidia-gpu", 0,error_elpa)
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)   
+   call e_complex%set("nvidia-gpu", TEST_GPU,error_elpa)
+#endif
+#if TEST_AMD_GPU == 1
+   call e_complex%set("amd-gpu", TEST_GPU,error_elpa)
+#endif
+
    call e_complex%set("omp_threads", 8, error_elpa)
 
    assert_elpa_ok(e_complex%setup())
@@ -271,7 +285,13 @@ program test
    call e_skewsymmetric%set("timings",1, error_elpa)
 
    call e_skewsymmetric%set("debug",1,error_elpa)
-   call e_skewsymmetric%set("nvidia-gpu", 0,error_elpa)
+
+#if TEST_NVIDIA_GPU == 1 || (TEST_NVIDIA_GPU == 0) && (TEST_AMD_GPU == 0)
+   call e_skewsymmetric%set("nvidia-gpu", TEST_GPU,error_elpa)
+#endif
+#if TEST_AMD_GPU == 1
+   call e_skewsymmetric%set("amd-gpu", TEST_GPU,error_elpa)
+#endif
    call e_skewsymmetric%set("omp_threads",8, error_elpa)
 
    assert_elpa_ok(e_skewsymmetric%setup())