Marked non-templated function definitions inline

This is to avoid multiple definitions compilation error when using
multiple translation units.

Author: oleksandr-pavlyk

dpctl/apis/include/dpctl4pybind11.hpp

@@ -372,9 +372,10 @@ class usm_memory : public py::object
 namespace tensor
 {
 
-std::vector<py::ssize_t> c_contiguous_strides(int nd,
-                                              const py::ssize_t *shape,
-                                              py::ssize_t element_size = 1)
+inline std::vector<py::ssize_t>
+c_contiguous_strides(int nd,
+                     const py::ssize_t *shape,
+                     py::ssize_t element_size = 1)
 {
     if (nd > 0) {
         std::vector<py::ssize_t> c_strides(nd, element_size);
@@ -389,9 +390,10 @@ std::vector<py::ssize_t> c_contiguous_strides(int nd,
     }
 }
 
-std::vector<py::ssize_t> f_contiguous_strides(int nd,
-                                              const py::ssize_t *shape,
-                                              py::ssize_t element_size = 1)
+inline std::vector<py::ssize_t>
+f_contiguous_strides(int nd,
+                     const py::ssize_t *shape,
+                     py::ssize_t element_size = 1)
 {
     if (nd > 0) {
         std::vector<py::ssize_t> f_strides(nd, element_size);
@@ -406,14 +408,14 @@ std::vector<py::ssize_t> f_contiguous_strides(int nd,
     }
 }
 
-std::vector<py::ssize_t>
+inline std::vector<py::ssize_t>
 c_contiguous_strides(const std::vector<py::ssize_t> &shape,
                      py::ssize_t element_size = 1)
 {
     return c_contiguous_strides(shape.size(), shape.data(), element_size);
 }
 
-std::vector<py::ssize_t>
+inline std::vector<py::ssize_t>
 f_contiguous_strides(const std::vector<py::ssize_t> &shape,
                      py::ssize_t element_size = 1)
 {

dpctl/apis/include/dpctl_capi.h

@@ -48,7 +48,7 @@
  * shared objects defining this symbols, if they call `import_dpctl()`
  * prior to using those symbols.
  */
-void import_dpctl(void)
+static inline void import_dpctl(void)
 {
     import_dpctl___sycl_device();
     import_dpctl___sycl_context();

dpctl/tensor/libtensor/include/kernels/constructors.hpp

@@ -52,7 +52,7 @@ template <typename T> T unbox_py_scalar(py::object o)
     return py::cast<T>(o);
 }
 
-template <> sycl::half unbox_py_scalar<sycl::half>(py::object o)
+template <> inline sycl::half unbox_py_scalar<sycl::half>(py::object o)
 {
     float tmp = py::cast<float>(o);
     return static_cast<sycl::half>(tmp);