Fix openmp flags

setup.py

@@ -53,6 +53,13 @@ def write_version_file():
 ]
 
 
+def append_flags(flags, flags_to_append):
+    for flag in flags_to_append:
+        if not flag in flags:
+            flags.append(flag)
+    return flags
+
+
 def get_extensions():
     build_cuda = torch.cuda.is_available() or os.getenv('FORCE_CUDA',
                                                         '0') == '1'
@@ -71,9 +78,12 @@ def get_extensions():
             nvcc_flags = []
         else:
             nvcc_flags = nvcc_flags.split(' ')
-        for flag in ['--expt-extended-lambda', '-Xcompiler', '-fopenmp']:
-            if not flag in nvcc_flags:
-                nvcc_flags.append(flag)
+        nvcc_flags = append_flags(nvcc_flags, ['--expt-extended-lambda', '-Xcompiler'])
+        if openmp:
+            if sys.platform == 'linux':
+                nvcc_flags = append_flags(nvcc_flags, ['-fopenmp'])
+            elif sys.platform == 'win32':
+                nvcc_flags = append_flags(nvcc_flags, ['/openmp'])
         extra_compile_args = {
             'cxx': [],
             'nvcc': nvcc_flags,
@@ -86,10 +96,11 @@ def get_extensions():
             cxx_flags = []
         else:
             cxx_flags = cxx_flags.split(' ')
-        if sys.platform == 'linux' and openmp:
-            for flag in ['-fopenmp']:
-                if not flag in cxx_flags:
-                    cxx_flags.append(flag)
+        if openmp:
+            if sys.platform == 'linux':
+                cxx_flags = append_flags(cxx_flags, ['-fopenmp'])
+            elif sys.platform == 'win32':
+                cxx_flags = append_flags(cxx_flags, ['/openmp'])
         extra_compile_args = {
             'cxx': cxx_flags
         }