Make pybind11 modules GIL-free

Author: ndgrigorian

dpctl/tensor/libtensor/source/tensor_accumulation.cpp

@@ -29,7 +29,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_accumulation_impl, m)
+PYBIND11_MODULE(_tensor_accumulation_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_accumulator_functions(m);
 }

dpctl/tensor/libtensor/source/tensor_ctors.cpp

@@ -170,7 +170,7 @@ void init_dispatch_vectors(void)
 
 } // namespace
 
-PYBIND11_MODULE(_tensor_impl, m)
+PYBIND11_MODULE(_tensor_impl, m, py::mod_gil_not_used())
 {
     init_dispatch_tables();
     init_dispatch_vectors();

dpctl/tensor/libtensor/source/tensor_elementwise.cpp

@@ -28,7 +28,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_elementwise_impl, m)
+PYBIND11_MODULE(_tensor_elementwise_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_elementwise_functions(m);
 }

dpctl/tensor/libtensor/source/tensor_linalg.cpp

@@ -28,7 +28,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_linalg_impl, m)
+PYBIND11_MODULE(_tensor_linalg_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_dot(m);
 }

dpctl/tensor/libtensor/source/tensor_reductions.cpp

@@ -29,7 +29,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_reductions_impl, m)
+PYBIND11_MODULE(_tensor_reductions_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_reduction_functions(m);
 }

dpctl/tensor/libtensor/source/tensor_sorting.cpp

@@ -34,7 +34,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_sorting_impl, m)
+PYBIND11_MODULE(_tensor_sorting_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_merge_sort_functions(m);
     dpctl::tensor::py_internal::init_merge_argsort_functions(m);

dpctl/tensor/libtensor/source/tensor_sorting_radix.cpp

@@ -30,7 +30,7 @@
 
 namespace py = pybind11;
 
-PYBIND11_MODULE(_tensor_sorting_radix_impl, m)
+PYBIND11_MODULE(_tensor_sorting_radix_impl, m, py::mod_gil_not_used())
 {
     dpctl::tensor::py_internal::init_radix_sort_functions(m);
     dpctl::tensor::py_internal::init_radix_argsort_functions(m);

dpctl/utils/src/device_queries.cpp

@@ -132,7 +132,7 @@ std::uint32_t py_intel_memory_bus_width(const sycl::device &d)
 
 }; // namespace
 
-PYBIND11_MODULE(_device_queries, m)
+PYBIND11_MODULE(_device_queries, m, py::mod_gil_not_used())
 {
     m.def("intel_device_info_device_id", &py_intel_device_id,
           "Get ext_intel_device_id for the device, zero if not an intel device",

dpctl/utils/src/order_keeper.cpp

@@ -6,7 +6,7 @@
 #include "sequential_order_keeper.hpp"
 #include <sycl/sycl.hpp>
 
-PYBIND11_MODULE(_seq_order_keeper, m)
+PYBIND11_MODULE(_seq_order_keeper, m, py::mod_gil_not_used())
 {
     py::class_<SequentialOrder>(m, "_OrderManager")
         .def(py::init<std::size_t>())