hpx: overhaul of the package recipe (#11194)

* added latest versions + malloc variant
* added a variant to activate different instrumentations
* added a variant to activate or not networking
* added a variant to activate or not cuda and tools
* enforce c++ standard and make it consistent with boost

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

@@ -7,17 +7,122 @@
 from spack import *
 
 
-class Hpx(CMakePackage):
+class Hpx(CMakePackage, CudaPackage):
     """C++ runtime system for parallel and distributed applications."""
 
     homepage = "http://stellar.cct.lsu.edu/tag/hpx/"
-    url      = "http://stellar.cct.lsu.edu/files/hpx_1.0.0.tar.gz"
+    url = "https://github.com/STEllAR-GROUP/hpx/archive/1.2.1.tar.gz"
 
-    version('1.0.0', '4983e7c6402417ec794d40343e36e417')
+    version('1.2.1', sha256='8cba9b48e919035176d3b7bbfc2c110df6f07803256626f1dad8d9dde16ab77a')
+    version('1.2.0', sha256='20942314bd90064d9775f63b0e58a8ea146af5260a4c84d0854f9f968077c170')
+    version('1.1.0', sha256='1f28bbe58d8f0da600d60c3a74a644d75ac777b20a018a5c1c6030a470e8a1c9')
 
-    depends_on('boost@1.55.0:')
+    version('1.0.0', '4983e7c6402417ec794d40343e36e417', url='http://stellar.cct.lsu.edu/files/hpx_1.0.0')
-    depends_on('hwloc@1.6:')
+
+    variant('cxxstd',
+            default='17',
+            values=('98', '11', '14', '17'),
+            description='Use the specified C++ standard when building.')
+
+    variant(
+        'malloc', default='tcmalloc',
+        description='Define which allocator will be linked in',
+        values=('system', 'tcmalloc', 'jemalloc', 'tbbmalloc')
+    )
+
+    variant('instrumentation', values=any_combination_of(
+        'apex', 'google_perftools', 'papi', 'valgrind'
+    ), description='Add support for various kind of instrumentation')
+
+    variant('networking', default=True,
+            description='Support for networking and multi=node runs')
+    variant('tools', default=False, description='Build HPX tools')
+
+    depends_on('boost')
+    depends_on('hwloc')
+    depends_on('python', type=('build', 'test', 'run'))
+    depends_on('pkgconfig', type='build')
+    depends_on('git', type='build')
+
+    # Recommended dependency versions for 1.2.X
+    depends_on('cmake@3.9.0:', when='@:1.2.1', type='build')
+    depends_on('boost@1.62.0:', when='@:1.2.1')
+    depends_on('hwloc@1.11:', when='@:1.2.1')
+
+    # Recommended dependency versions before 1.2
+    depends_on('boost@1.55.0:', when='@:1.1.0')
+    depends_on('hwloc@1.6:', when='@:1.1.0')
+
+    # CXX Standard
+    depends_on('boost cxxstd=98', when='cxxstd=98')
+    depends_on('boost cxxstd=11', when='cxxstd=11')
+    depends_on('boost cxxstd=14', when='cxxstd=14')
+    depends_on('boost cxxstd=17', when='cxxstd=17')
+
+    # Malloc
+    depends_on('gperftools', when='malloc=tcmalloc')
+    depends_on('jemalloc', when='malloc=jemalloc')
+    depends_on('tbb', when='malloc=tbbmalloc')
+
+    # Instrumentation
+    depends_on('apex', when='instrumentation=apex')
+    depends_on('gperftools', when='instrumentation=google_perftools')
+    depends_on('papi', when='instrumentation=papi')
+    depends_on('valgrind', when='instrumentation=valgrind')
+
+    # TODO: hpx can build perfectly fine in parallel, except that each
+    # TODO: process might need more than 2GB to compile. This is just the
+    # TODO: most conservative approach to ensure a sane build.
+    parallel = False
+
+    def cxx_standard(self):
+        value = self.spec.variants['cxxstd'].value
+        value = '0X' if value == '98' else value
+        return '-DHPX_WITH_CXX{0}=ON'.format(value)
+
+    def instrumentation_args(self):
+        for value in self.variants['instrumentation'].values:
+            if value == 'none':
+                continue
+
+            condition = 'instrumentation={0}'.format(value)
+            yield '-DHPX_WITH_{0}={1}'.format(
+                str(value).upper(), 'ON' if condition in self.spec else 'OFF'
+            )
 
     def cmake_args(self):
-        args = ['-DHPX_BUILD_EXAMPLES=OFF', '-DHPX_MALLOC=system']
+        spec, args = self.spec, []
+
+        # CXX Standard
+        args.append(self.cxx_standard())
+
+        # Malloc
+        selected_malloc = spec.variants['malloc'].value
+        args.append('-DHPX_WITH_MALLOC={0}'.format(selected_malloc))
+
+        # Instrumentation
+        args.extend(self.instrumentation_args())
+
+        # Networking
+        args.append('-DHPX_WITH_NETWORKING={0}'.format(
+            'ON' if '+networking' in spec else 'OFF'
+        ))
+
+        # Cuda support
+        args.append('-DHPX_WITH_CUDA={0}'.format(
+            'ON' if '+cuda' in spec else 'OFF'
+        ))
+
+        # Tools
+        args.append('-DHPX_WITH_TOOLS={0}'.format(
+            'ON' if '+tools' in spec else 'OFF'
+        ))
+
+        args.extend([
+            '-DBOOST_ROOT={0}'.format(spec['boost'].prefix),
+            '-DHWLOC_ROOT={0}'.format(spec['hwloc'].prefix),
+            '-DHPX_WITH_BOOST_ALL_DYNAMIC_LINK=ON',
+            '-DBUILD_SHARED_LIBS=ON'
+        ])
+
         return args