Merge move_tensor_impl_ext into move_tensor_impl_ext_part_2

Author: vlad-perevezentsev

.gitignore

@@ -33,4 +33,6 @@ dpnp/**/*.pyd
 *~
 core
 
+# TODO: revert to `dpctl/`
+# when dpnp fully migrates dpctl/tensor
 dpctl_ext/**/*.cpython*.so

dpctl_ext/tensor/CMakeLists.txt

@@ -27,7 +27,7 @@
 # THE POSSIBILITY OF SUCH DAMAGE.
 # *****************************************************************************
 
-find_package(Python COMPONENTS Development)
+find_package(Python COMPONENTS Development.Module)
 
 if(WIN32)
     if(${CMAKE_VERSION} VERSION_LESS "3.27")
@@ -74,7 +74,7 @@ target_include_directories(
         # ${Dpctl_INCLUDE_DIR}
         ${CMAKE_CURRENT_SOURCE_DIR}/libtensor/include
 )
-target_link_libraries(${_static_lib_trgt} PRIVATE pybind11::headers Python::Python)
+target_link_libraries(${_static_lib_trgt} PRIVATE pybind11::headers Python::Module)
 set_target_properties(${_static_lib_trgt} PROPERTIES POSITION_INDEPENDENT_CODE ON)
 
 set(_py_trgts)

dpctl_ext/tensor/libtensor/source/copy_as_contig.cpp

@@ -189,6 +189,12 @@ std::pair<sycl::event, sycl::event>
             "Execution queue is not compatible with allocation queues");
     }
 
+    // check that arrays do not overlap, and concurrent copying is safe.
+    auto const &overlap = dpctl::tensor::overlap::MemoryOverlap();
+    if (overlap(src, dst)) {
+        throw py::value_error("Arrays index overlapping segments of memory");
+    }
+
     const auto &src_strides_vec = src.get_strides_vector();
 
     if (src_nd >= 2) {
@@ -314,6 +320,12 @@ std::pair<sycl::event, sycl::event>
             "Execution queue is not compatible with allocation queues");
     }
 
+    // check that arrays do not overlap, and concurrent copying is safe.
+    auto const &overlap = dpctl::tensor::overlap::MemoryOverlap();
+    if (overlap(src, dst)) {
+        throw py::value_error("Arrays index overlapping segments of memory");
+    }
+
     const auto &src_strides_vec = src.get_strides_vector();
 
     if (src_nd >= 2) {
@@ -459,6 +471,12 @@ std::pair<sycl::event, sycl::event>
             "Execution queue is not compatible with allocation queues");
     }
 
+    // check that arrays do not overlap, and concurrent copying is safe.
+    auto const &overlap = dpctl::tensor::overlap::MemoryOverlap();
+    if (overlap(src, dst)) {
+        throw py::value_error("Arrays index overlapping segments of memory");
+    }
+
     if (nelems == 0) {
         // nothing to do
         return std::make_pair(sycl::event(), sycl::event());
@@ -624,6 +642,20 @@ std::pair<sycl::event, sycl::event>
         throw py::value_error("Unexpected destination array layout");
     }
 
+    dpctl::tensor::validation::CheckWritable::throw_if_not_writable(dst);
+
+    // check compatibility of execution queue and allocation queue
+    if (!dpctl::utils::queues_are_compatible(exec_q, {src, dst})) {
+        throw py::value_error(
+            "Execution queue is not compatible with allocation queues");
+    }
+
+    // check that arrays do not overlap, and concurrent copying is safe.
+    auto const &overlap = dpctl::tensor::overlap::MemoryOverlap();
+    if (overlap(src, dst)) {
+        throw py::value_error("Arrays index overlapping segments of memory");
+    }
+
     int src_typenum = src.get_typenum();
     int dst_typenum = dst.get_typenum();
 

dpctl_ext/tensor/libtensor/source/tensor_ctors.cpp

@@ -32,18 +32,15 @@
 /// This file defines functions of dpctl.tensor._tensor_impl extensions
 //===----------------------------------------------------------------------===//
 
-// #include <algorithm>
-// #include <complex>
-// #include <cstdint>
-#include <pybind11/pybind11.h>
-#include <pybind11/stl.h>
-#include <sycl/sycl.hpp>
-// #include <thread>
 #include <optional>
 #include <type_traits>
 #include <utility>
 #include <vector>
 
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+#include <sycl/sycl.hpp>
+
 #include "dpnp4pybind11.hpp"
 
 // #include "accumulators.hpp"