sphexa package (#48128)

* sphexa package

* remove older versions

* avoid setting args twice

Co-authored-by: Rocco Meli <r.meli@bluemail.ch>

* rocprim should be hipcub

* address review comments

---------

Co-authored-by: Rocco Meli <r.meli@bluemail.ch>

var/spack/repos/builtin/packages/sphexa/package.py

@@ -0,0 +1,77 @@
+# Copyright 2013-2024 Lawrence Livermore National Security, LLC and other
+# Spack Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: (Apache-2.0 OR MIT)
+
+
+from spack.package import *
+
+
+class Sphexa(CMakePackage, CudaPackage, ROCmPackage):
+    """SPH and N-body simulation framework"""
+
+    homepage = "https://github.com/sphexa-org/sphexa"
+    url = "https://github.com/sphexa-org/sphexa/archive/v0.0.0.tar.gz"
+    git = "https://github.com/sphexa-org/sphexa.git"
+    maintainers = ["sekelle"]
+
+    license("MIT")
+
+    version("0.93.1", sha256="95a93d0063ac8857b9be12c1aca24f5b2eef9dd4ffe8cf3f6b552a4dd54b940f")
+    version("develop", branch="develop")
+
+    variant("hdf5", default=True, description="Enable support for HDF5 I/O")
+    variant("gpu_aware_mpi", default=True, description="GPU aware MPI")
+
+    depends_on("cmake@3.22:", type="build")
+
+    depends_on("mpi")
+    depends_on("cuda@11.2:", when="+cuda")
+    depends_on("hip", when="+rocm")
+    depends_on("rocthrust", when="+rocm")
+    depends_on("hipcub", when="+rocm")
+    depends_on("hdf5 +mpi", when="+hdf5")
+
+    # Build MPI with GPU support when GPU aware MPI is requested.
+    # For cray-mpich, the user is responsible to configure it for GPU aware MPI.
+    with when("+gpu_aware_mpi"):
+        depends_on("openmpi +cuda", when="+cuda ^[virtuals=mpi] openmpi")
+        depends_on("mpich +cuda", when="+cuda ^[virtuals=mpi] mpich")
+        depends_on("mvapich +cuda", when="+cuda ^[virtuals=mpi] mvapich")
+        depends_on("mvapich2 +cuda", when="+cuda ^[virtuals=mpi] mvapich2")
+
+        depends_on("mpich +rocm", when="+rocm ^[virtuals=mpi] mpich")
+
+    conflicts("%gcc@:10")
+    conflicts("cuda_arch=none", when="+cuda", msg="CUDA architecture is required")
+    conflicts("amdgpu_target=none", when="+rocm", msg="HIP architecture is required")
+    conflicts("+cuda", when="+rocm", msg="CUDA and HIP cannot both be enabled")
+
+    def cmake_args(self):
+        spec = self.spec
+
+        args = [
+            self.define_from_variant("SPH_EXA_WITH_H5PART", "hdf5"),
+            self.define_from_variant("SPH_EXA_WITH_CUDA", "cuda"),
+            self.define_from_variant("RYOANJI_WITH_CUDA", "cuda"),
+            self.define_from_variant("CSTONE_WITH_CUDA", "cuda"),
+            self.define_from_variant("SPH_EXA_WITH_HIP", "rocm"),
+            self.define_from_variant("RYOANJI_WITH_HIP", "rocm"),
+            self.define_from_variant("CSTONE_WITH_HIP", "rocm"),
+        ]
+
+        if spec.satisfies("+rocm") or spec.satisfies("+cuda"):
+            args.append(self.define_from_variant("CSTONE_WITH_GPU_AWARE_MPI", "gpu_aware_mpi"))
+
+        if spec.satisfies("+rocm"):
+            archs = spec.variants["amdgpu_target"].value
+            arch_str = ";".join(archs)
+            args.append(self.define("CMAKE_HIP_ARCHITECTURES", arch_str))
+
+        if spec.satisfies("+cuda"):
+            args.append(self.define("CMAKE_CUDA_FLAGS", "-ccbin={0}".format(spec["mpi"].mpicxx)))
+            archs = spec.variants["cuda_arch"].value
+            arch_str = ";".join(archs)
+            args.append(self.define("CMAKE_CUDA_ARCHITECTURES", arch_str))
+
+        return args