Added HIP support to VTK-m package (#19523)

var/spack/repos/builtin/packages/vtk-m/package.py

@@ -40,32 +40,52 @@ class VtkM(CMakePackage, CudaPackage):
     variant('build_type', default='Release', description='CMake build type',
             values=('Debug', 'Release', 'RelWithDebInfo', 'MinSizeRel'))
     variant("shared", default=False, description="build shared libs")
-    variant("cuda", default=False, description="build cuda support")
+
     variant("doubleprecision", default=True,
             description='enable double precision')
     variant("logging", default=False, description="build logging support")
     variant("mpi", default=False, description="build mpi support")
-    variant("openmp", default=(sys.platform != 'darwin'), description="build openmp support")
     variant("rendering", default=True, description="build rendering support")
-    variant("tbb", default=(sys.platform == 'darwin'), description="build TBB support")
     variant("64bitids", default=False,
             description="enable 64 bits ids")
 
+    # Device variants
+    variant("cuda", default=False, description="build cuda support")
+    variant("openmp", default=(sys.platform != 'darwin'), description="build openmp support")
+    variant("tbb", default=(sys.platform == 'darwin'), description="build TBB support")
+    variant("hip", default=False, description="build hip support")
+
+    # it doesn't look like spack has a amd gpu abstraction
+    amdgpu_targets = (
+        'gfx701', 'gfx801', 'gfx802', 'gfx803',
+        'gfx900', 'gfx906', 'gfx908', 'gfx1010',
+        'gfx1011', 'gfx1012'
+    )
+    variant('amdgpu_target', default='none', multi=True, values=amdgpu_targets)
+    conflicts("+hip", when="amdgpu_target=none")
+
     depends_on("cmake@3.12:", type="build")         # CMake >= 3.12
+    depends_on("cmake@3.18:", when="+hip")          # CMake >= 3.18
+
+    depends_on('cuda@10.2.0:', when='+cuda')
     depends_on("tbb", when="+tbb")
-    depends_on("cuda", when="+cuda")
     depends_on("mpi", when="+mpi")
 
+    depends_on("kokkos@3.1:+hip", when="+hip")
+    depends_on("rocm-cmake@3.7:", when="+hip")
+    depends_on("hip@3.7:", when="+hip")
+
+    conflicts("+hip", when="+cuda")
     conflicts("~shared", when="~pic")
 
     def cmake_args(self):
         spec = self.spec
         options = []
-        gpu_name_table = {'20': 'fermi',
+        gpu_name_table = {'30': 'kepler',  '32': 'kepler',  '35': 'kepler',
-                          '30': 'kepler',  '32': 'kepler',  '35': 'kepler',
                           '50': 'maxwell', '52': 'maxwell', '53': 'maxwell',
                           '60': 'pascal',  '61': 'pascal',  '62': 'pascal',
-                          '70': 'volta',   '72': 'turing',  '75': 'turing'}
+                          '70': 'volta',   '72': 'turing',  '75': 'turing',
+                          '80': 'ampere',  '86': 'ampere'}
         with working_dir('spack-build', create=True):
             options = ["-DVTKm_ENABLE_TESTING:BOOL=OFF"]
             # shared vs static libs logic
@@ -77,26 +97,6 @@ def cmake_args(self):
                     options.append('-DBUILD_SHARED_LIBS=ON')
                 else:
                     options.append('-DBUILD_SHARED_LIBS=OFF')
-            # cuda support
-            if "+cuda" in spec:
-                options.append("-DVTKm_ENABLE_CUDA:BOOL=ON")
-                options.append("-DCMAKE_CUDA_HOST_COMPILER={0}".format(
-                               env["SPACK_CXX"]))
-                if 'cuda_arch' in spec.variants:
-                    cuda_value = spec.variants['cuda_arch'].value
-                    cuda_arch = cuda_value[0]
-                    if cuda_arch in gpu_name_table:
-                        vtkm_cuda_arch = gpu_name_table[cuda_arch]
-                        options.append('-DVTKm_CUDA_Architecture={0}'.format(
-                                       vtkm_cuda_arch))
-                else:
-                    # this fix is necessary if compiling platform has cuda, but
-                    # no devices (this's common for front end nodes on hpc clus
-                    # ters)
-                    # we choose kepler as a lowest common denominator
-                    options.append("-DVTKm_CUDA_Architecture=kepler")
-            else:
-                options.append("-DVTKm_ENABLE_CUDA:BOOL=OFF")
 
             # double precision
             if "+doubleprecision" in spec:
@@ -122,30 +122,12 @@ def cmake_args(self):
             else:
                 options.append("-DVTKm_ENABLE_MPI:BOOL=OFF")
 
-            # openmp support
-            if "+openmp" in spec:
-                # openmp is added since version 1.3.0
-                if not spec.satisfies('@1.3.0:,ascent_ver'):
-                    raise InstallError('OpenMP is not supported for\
-                            vtkm version lower than 1.3')
-                options.append("-DVTKm_ENABLE_OPENMP:BOOL=ON")
-            else:
-                options.append("-DVTKm_ENABLE_OPENMP:BOOL=OFF")
-
             # rendering support
             if "+rendering" in spec:
                 options.append("-DVTKm_ENABLE_RENDERING:BOOL=ON")
             else:
                 options.append("-DVTKm_ENABLE_RENDERING:BOOL=OFF")
 
-            # tbb support
-            if "+tbb" in spec:
-                # vtk-m detectes tbb via TBB_ROOT env var
-                os.environ["TBB_ROOT"] = spec["tbb"].prefix
-                options.append("-DVTKm_ENABLE_TBB:BOOL=ON")
-            else:
-                options.append("-DVTKm_ENABLE_TBB:BOOL=OFF")
-
             # 64 bit ids
             if "+64bitids" in spec:
                 options.append("-DVTKm_USE_64BIT_IDS:BOOL=ON")
@@ -156,4 +138,52 @@ def cmake_args(self):
             if spec.variants["build_type"].value != 'Release':
                 options.append("-DVTKm_NO_ASSERT:BOOL=ON")
 
+            # cuda support
+            if "+cuda" in spec:
+                options.append("-DVTKm_ENABLE_CUDA:BOOL=ON")
+                options.append("-DCMAKE_CUDA_HOST_COMPILER={0}".format(
+                               env["SPACK_CXX"]))
+                if 'cuda_arch' in spec.variants:
+                    cuda_value = spec.variants['cuda_arch'].value
+                    cuda_arch = cuda_value[0]
+                    if cuda_arch in gpu_name_table:
+                        vtkm_cuda_arch = gpu_name_table[cuda_arch]
+                        options.append('-DVTKm_CUDA_Architecture={0}'.format(
+                                       vtkm_cuda_arch))
+                else:
+                    # this fix is necessary if compiling platform has cuda, but
+                    # no devices (this is common for front end nodes on hpc
+                    # clusters). We choose volta as a lowest common denominator
+                    options.append("-DVTKm_CUDA_Architecture=volta")
+            else:
+                options.append("-DVTKm_ENABLE_CUDA:BOOL=OFF")
+
+            # hip support
+            if "+hip" in spec:
+                options.append("-DVTKm_ENABLE_HIP:BOOL=ON")
+
+                archs = ",".join(self.spec.variants['amdgpu_target'].value)
+                options.append(
+                    "-DCMAKE_HIP_ARCHITECTURES:STRING={0}".format(archs))
+            else:
+                options.append("-DVTKm_ENABLE_HIP:BOOL=OFF")
+
+            # openmp support
+            if "+openmp" in spec:
+                # openmp is added since version 1.3.0
+                if not spec.satisfies('@1.3.0:,ascent_ver'):
+                    raise InstallError('OpenMP is not supported for\
+                            vtkm version lower than 1.3')
+                options.append("-DVTKm_ENABLE_OPENMP:BOOL=ON")
+            else:
+                options.append("-DVTKm_ENABLE_OPENMP:BOOL=OFF")
+
+            # tbb support
+            if "+tbb" in spec:
+                # vtk-m detectes tbb via TBB_ROOT env var
+                os.environ["TBB_ROOT"] = spec["tbb"].prefix
+                options.append("-DVTKm_ENABLE_TBB:BOOL=ON")
+            else:
+                options.append("-DVTKm_ENABLE_TBB:BOOL=OFF")
+
             return options