Test commit splitting up elementwise functions

Author: ndgrigorian

dpctl/tensor/CMakeLists.txt

@@ -47,7 +47,8 @@ set(_tensor_impl_sources
     ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/where.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/boolean_reductions.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/device_support_queries.cpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions.cpp
+    ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions1.cpp
+    ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions2.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/repeat.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/reduction_over_axis.cpp
 )
@@ -63,7 +64,8 @@ endif()
 set(_no_fast_math_sources
   ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/full_ctor.cpp
   ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/linear_sequences.cpp
-  ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions.cpp
+  ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions1.cpp
+  ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/elementwise_functions2.cpp
   ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/source/reduction_over_axis.cpp
 )
 foreach(_src_fn ${_no_fast_math_sources})

dpctl/tensor/libtensor/source/elementwise_functions.hpp

@@ -22,7 +22,6 @@
 /// This file defines functions of dpctl.tensor._tensor_impl extensions,
 /// specifically functions for elementwise operations.
 //===----------------------------------------------------------------------===//
-
 #pragma once
 
 #include "dpctl4pybind11.hpp"
@@ -38,18 +37,61 @@
 #include "utils/offset_utils.hpp"
 #include "utils/type_dispatch.hpp"
 
+namespace td_ns = dpctl::tensor::type_dispatch;
+
+static inline py::dtype _dtype_from_typenum(td_ns::typenum_t dst_typenum_t)
+{
+    switch (dst_typenum_t) {
+    case td_ns::typenum_t::BOOL:
+        return py::dtype("?");
+    case td_ns::typenum_t::INT8:
+        return py::dtype("i1");
+    case td_ns::typenum_t::UINT8:
+        return py::dtype("u1");
+    case td_ns::typenum_t::INT16:
+        return py::dtype("i2");
+    case td_ns::typenum_t::UINT16:
+        return py::dtype("u2");
+    case td_ns::typenum_t::INT32:
+        return py::dtype("i4");
+    case td_ns::typenum_t::UINT32:
+        return py::dtype("u4");
+    case td_ns::typenum_t::INT64:
+        return py::dtype("i8");
+    case td_ns::typenum_t::UINT64:
+        return py::dtype("u8");
+    case td_ns::typenum_t::HALF:
+        return py::dtype("f2");
+    case td_ns::typenum_t::FLOAT:
+        return py::dtype("f4");
+    case td_ns::typenum_t::DOUBLE:
+        return py::dtype("f8");
+    case td_ns::typenum_t::CFLOAT:
+        return py::dtype("c8");
+    case td_ns::typenum_t::CDOUBLE:
+        return py::dtype("c16");
+    default:
+        throw py::value_error("Unrecognized dst_typeid");
+    }
+}
+
+static inline int _result_typeid(int arg_typeid, const int *fn_output_id)
+{
+    if (arg_typeid < 0 || arg_typeid >= td_ns::num_types) {
+        throw py::value_error("Input typeid " + std::to_string(arg_typeid) +
+                              " is outside of expected bounds.");
+    }
+
+    return fn_output_id[arg_typeid];
+}
+
 namespace dpctl
 {
 namespace tensor
 {
 namespace py_internal
 {
 
-namespace td_ns = dpctl::tensor::type_dispatch;
-
-extern py::dtype _dtype_from_typenum(td_ns::typenum_t dst_typenum_t);
-extern int _result_typeid(int arg_typeid, const int *fn_output_id);
-
 template <typename output_typesT,
           typename contig_dispatchT,
           typename strided_dispatchT>
@@ -825,8 +867,6 @@ py_binary_inplace_ufunc(const dpctl::tensor::usm_ndarray &lhs,
         strided_fn_ev);
 }
 
-extern void init_elementwise_functions(py::module_ m);
-
 } // namespace py_internal
 } // namespace tensor
 } // namespace dpctl