neuron: add new versions and clean-up the recipe (#41931)

* update tarball urls, add new versions and update maintainers
* remove unnecessary variant like cross-compile
* use self.define and self.define_from_variant
* improve regex / bug that matches with -DMPICH_SKIP_MPICXX=1

Co-authored-by: Matthias Wolf <matthias.wolf@epfl.ch>
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>

1 files changed, 108 insertions, 66 deletions

diff --git a/var/spack/repos/builtin/packages/neuron/package.py b/var/spack/repos/builtin/packages/neuron/package.py
index edad5ec994..c6d33e7030 100644
--- a/var/spack/repos/builtin/packages/neuron/package.py
+++ b/var/spack/repos/builtin/packages/neuron/package.py
@@ -16,86 +16,137 @@ class Neuron(CMakePackage):
     """
 
     homepage = "https://www.neuron.yale.edu/"
-    url = "https://neuron.yale.edu/ftp/neuron/versions/v7.7/nrn-7.7.tar.gz"
+    url = "https://github.com/neuronsimulator/nrn/releases/download/8.2.3/nrn-full-src-package-8.2.3.tar.gz"
     git = "https://github.com/neuronsimulator/nrn"
-    maintainers("pramodk", "nrnhines", "iomaganaris", "alexsavulescu")
+    maintainers("pramodk", "nrnhines", "iomaganaris", "ohm314", "matz-e")
 
     license("BSD-3-Clause")
 
     version("develop", branch="master", submodules="True")
+
     version(
-        "8.0.0", tag="8.0.0", commit="429d11ef34b1d860b3ddbfffc9f7960acb399b0c", submodules="True"
+        "8.2.3", tag="8.2.3", commit="f0ed3701059aa53ce93387f3d73d13c45de6d87f", submodules="True"
     )
     version(
-        "7.8.2", tag="7.8.2", commit="09b151ecb2b3984335c265932dc6ba3e4fcb318e", submodules="True"
+        "8.1.0", tag="8.1.0", commit="047dd8240c2badadf5ea154b563b29369db1303f", submodules="True"
+    )
+    version(
+        "8.0.0", tag="8.0.0", commit="429d11ef34b1d860b3ddbfffc9f7960acb399b0c", submodules="True"
     )
     version(
-        "7.8.1", tag="7.8.1", commit="47cd8c85aa3fde5dbb7319facd6f475438235d7d", submodules="True"
+        "7.8.2", tag="7.8.2", commit="09b151ecb2b3984335c265932dc6ba3e4fcb318e", submodules="True"
     )
 
-    variant("coreneuron", default=False, description="Enable CoreNEURON as submodule")
-    variant("cross-compile", default=False, description="Build for cross-compile environment")
+    variant("backtrace", default=False, description="Enable printing backtraces on failure")
     variant("interviews", default=False, description="Enable GUI with INTERVIEWS")
     variant("legacy-unit", default=False, description="Enable legacy units")
     variant("mpi", default=True, description="Enable MPI parallelism")
     variant("python", default=True, description="Enable python")
-    variant("rx3d", default=False, description="Enable cython translated 3-d rxd")
-    variant("tests", default=False, description="Enable unit tests")
-    variant("caliper", default=False, description="Add LLNL/Caliper support")
+    variant("shared", default=True, description="Build shared library (CoreNEURON)")
+    variant("tests", default=False, description="Enable building tests")
+    variant("rx3d", default=False, description="Enable cython translated 3-d rxd.", when="+python")
+
+    # variants from coreneuron support
+    variant("coreneuron", default=True, description="Enable CoreNEURON support")
+    variant(
+        "gpu", default=False, description="Enable GPU build (with NVHPC)", when="@9:+coreneuron"
+    )
+    variant(
+        "openmp", default=False, description="Enable CoreNEURON OpenMP support", when="+coreneuron"
+    )
+    variant(
+        "sympy",
+        default=False,
+        description="Use NMODL with SymPy to solve ODEs",
+        when="@9:+coreneuron",
+    )
+    variant("caliper", default=False, description="Add Caliper support")
 
-    depends_on("bison", type="build")
-    depends_on("flex", type="build")
-    depends_on("py-cython", when="+rx3d", type="build")
+    generator("ninja")
+
+    depends_on("bison@3:", type="build")
+    depends_on("flex@2.6:", type="build")
+    depends_on("ninja", type="build")
 
     depends_on("gettext")
+    depends_on("libdwarf", when="+backtrace")
     depends_on("mpi", when="+mpi")
     depends_on("ncurses")
-    depends_on("python@2.7:", when="+python")
+    depends_on("readline")
+
+    depends_on("python", when="+python")
     depends_on("py-pytest", when="+python+tests")
     depends_on("py-mpi4py", when="+mpi+python+tests")
-    depends_on("readline")
+    depends_on("py-numpy", when="+python")
+    depends_on("py-cython", when="+rx3d", type="build")
+    depends_on("py-pytest-cov", when="+tests")
+
+    # next two needed after neuronsimulator/nrn#2235.
+    depends_on("py-pip", type="build")
+    depends_on("py-setuptools", type="build")
+    depends_on("py-packaging", type="run")
+
+    depends_on("boost", when="+coreneuron+tests")
+    depends_on("cuda", when="+coreneuron+gpu")
+    depends_on("py-sympy@1.3:", when="+coreneuron")
+
     depends_on("caliper", when="+caliper")
-    depends_on("py-numpy", type="run")
 
-    conflicts("+rx3d", when="~python")
+    gpu_compiler_message = "neuron+gpu needs %nvhpc"
+    requires("%nvhpc", when="+gpu", msg=gpu_compiler_message)
 
     patch("patch-v782-git-cmake-avx512.patch", when="@7.8.2")
 
     def cmake_args(self):
         spec = self.spec
+        args = []
+        for variant in ["backtrace", "coreneuron", "interviews", "mpi", "python", "rx3d", "tests"]:
+            args.append(self.define_from_variant("NRN_ENABLE_" + variant.upper(), variant))
 
-        def cmake_options(spec_options):
-            value = "TRUE" if spec_options in spec else "FALSE"
-            cmake_name = spec_options[1:].upper().replace("-", "_")
-            return "-DNRN_ENABLE_" + cmake_name + ":BOOL=" + value
-
-        args = [
-            cmake_options(variant)
-            for variant in [
-                "+coreneuron",
-                "+interviews",
-                "+mpi",
-                "+python",
-                "+rx3d",
-                "+coreneuron",
-                "+tests",
-            ]
-        ]
-        args.append("-DNRN_ENABLE_BINARY_SPECIAL=ON")
+        args.append(self.define_from_variant("CORENRN_ENABLE_SHARED", "shared"))
 
-        if "~mpi" in spec and "+coreneuron" in spec:
-            args.append("-DCORENRN_ENABLE_MPI=OFF")
+        if spec.satisfies("@:8"):
+            args.append(self.define("NRN_ENABLE_BINARY_SPECIAL", "ON"))
 
-        if spec.variants["build_type"].value == "Debug":
-            args.append("-DCMAKE_C_FLAGS=-g -O0")
-            args.append("-DCMAKE_CXX_FLAGS=-g -O0")
-            args.append("-DCMAKE_BUILD_TYPE=Custom")
+        if "+python" in spec:
+            args.append(self.define("PYTHON_EXECUTABLE", spec["python"].command.path))
 
-        if "+legacy-unit" in spec:
-            args.append("-DNRN_DYNAMIC_UNITS_USE_LEGACY=ON")
+        if "+legacy-unit" in spec and spec.satisfies("@:8"):
+            args.append(self.define("NRN_DYNAMIC_UNITS_USE_LEGACY", "ON"))
 
         if "+caliper" in spec:
-            args.append("-DCORENRN_CALIPER_PROFILING=ON")
+            args.append(self.define("NRN_ENABLE_PROFILING", "ON"))
+            args.append(self.define("NRN_PROFILER", "caliper"))
+
+        if spec.satisfies("+coreneuron"):
+            options = [
+                self.define("CORENRN_ENABLE_SPLAYTREE_QUEUING", "ON"),
+                self.define("CORENRN_ENABLE_TIMEOUT", "OFF"),
+                self.define_from_variant("CORENRN_ENABLE_OPENMP", "openmp"),
+                self.define_from_variant("CORENRN_ENABLE_LEGACY_UNITS", "legacy-unit"),
+                self.define_from_variant("CORENRN_ENABLE_UNIT_TESTS", "tests"),
+            ]
+
+            nmodl_options = "codegen --force"
+            if spec.satisfies("+sympy"):
+                nmodl_options += " sympy --analytic"
+            options.append(self.define("CORENRN_NMODL_FLAGS", nmodl_options))
+
+            if spec.satisfies("+gpu"):
+                nvcc = spec["cuda"].prefix.bin.nvcc
+                options.append(self.define("CMAKE_CUDA_COMPILER", nvcc))
+                options.append(self.define("CORENRN_ENABLE_GPU", True))
+
+            args.extend(options)
+
+        # Enable math optimisations to enable SIMD/vectorisation in release modes
+        if spec.satisfies("build_type=Release") or spec.satisfies("build_type=RelWithDebInfo"):
+            args.append(self.define("NRN_ENABLE_MATH_OPT", "ON"))
+
+        # add cpu arch specific optimisation flags to CMake so that they are passed
+        # to embedded Makefile that neuron has for compiling MOD files
+        compilation_flags = self.spec.architecture.target.optimization_flags(self.spec.compiler)
+        args.append(self.define("CMAKE_CXX_FLAGS", compilation_flags))
 
         return args
 
@@ -117,30 +168,21 @@ class Neuron(CMakePackage):
             cxx_compiler = self.compiler.cxx
 
         kwargs = {"backup": False, "string": True}
-        nrnmech_makefile = join_path(self.prefix, "./bin/nrnmech_makefile")
+        nrnmech_makefile = join_path(self.prefix, "bin/nrnmech_makefile")
 
         # assign_operator is changed to fix wheel support
-        if self.spec.satisfies("@:7"):
-            assign_operator = "?="
-        else:
-            assign_operator = "="
-
-        filter_file(
-            "CC {0} {1}".format(assign_operator, env["CC"]),
-            "CC = {0}".format(cc_compiler),
-            nrnmech_makefile,
-            **kwargs,
-        )
-        filter_file(
-            "CXX {0} {1}".format(assign_operator, env["CXX"]),
-            "CXX = {0}".format(cxx_compiler),
-            nrnmech_makefile,
-            **kwargs,
-        )
-
-        if spec.satisfies("+coreneuron"):
-            corenrn_makefile = join_path(self.prefix, "share/coreneuron/nrnivmodl_core_makefile")
-            filter_file(env["CXX"], cxx_compiler, corenrn_makefile, **kwargs)
+        assign_operator = "?=" if spec.satisfies("@:7") else "="
+
+        # replace compilers from makefile
+        compilers = [("CC", "cc_compiler"), ("CXX", "cxx_compiler")]
+        for compiler_var, compiler_env in compilers:
+            pattern = "(?:^|\\s){0}\\s*{1}.+".format(compiler_var, assign_operator)
+            replacement = "{0} = {1}".format(compiler_var, locals()[compiler_env])
+            filter_file(pattern, replacement, nrnmech_makefile)
+
+        if spec.satisfies("@8:+coreneuron"):
+            nrnmakefile = join_path(self.prefix, "share/coreneuron/nrnivmodl_core_makefile")
+            filter_file("(?:^|\\s)CXX\\s*=.+", "CXX = {0}".format(cxx_compiler), nrnmakefile)
 
     def setup_run_environment(self, env):
         env.prepend_path("PATH", join_path(self.prefix, "bin"))