Merge branch 'fix_full_array_dtype' of https://github.com/IntelPython/dpctl into fix_full_array_dtype

Author: npolina4

.flake8

@@ -32,3 +32,4 @@ per-file-ignores =
     examples/cython/sycl_buffer/_buffer_example.pyx: E999, E225, E402
     examples/cython/sycl_direct_linkage/_buffer_example.pyx: E999, E225, E402
     examples/cython/usm_memory/blackscholes.pyx: E999, E225, E226, E402
+    examples/cython/use_dpctl_sycl/use_dpctl_sycl/_cython_api.pyx: E999, E225, E226, E402

.github/workflows/conda-package.yml

@@ -160,15 +160,13 @@ jobs:
           . $CONDA/etc/profile.d/conda.sh
           conda activate test_dpctl
           export OCL_ICD_FILENAMES=libintelocl.so
-          export SYCL_ENABLE_HOST_DEVICE=1
           python -c "import dpctl; dpctl.lsplatform()"
       - name: Run tests
         run: |
           . $CONDA/etc/profile.d/conda.sh
           conda activate test_dpctl
           # echo "libintelocl.so" | tee /etc/OpenCL/vendors/intel-cpu.icd
           export OCL_ICD_FILENAMES=libintelocl.so
-          export SYCL_ENABLE_HOST_DEVICE=1
           # clinfo -l
           python -m pytest -p no:faulthandler --pyargs $MODULE_NAME
 
@@ -297,14 +295,12 @@ jobs:
         shell: cmd /C CALL {0}
         run: |
           python -c "import sys; print(sys.executable)"
-          set SYCL_ENABLE_HOST_DEVICE=1
           python -c "import dpctl; dpctl.lsplatform()"
           python -c "import dpctl; print(dpctl.get_devices(backend='opencl', device_type='gpu'))"
           python -c "import dpctl; print(dpctl.get_num_devices(backend='opencl', device_type='gpu'))"
       - name: Run tests
         shell: cmd /C CALL {0}
         run: |
-          set SYCL_ENABLE_HOST_DEVICE=1
           python -c "import sys; print(sys.executable)"
           python -m pytest -p no:faulthandler --pyargs ${{ env.MODULE_NAME }}
 
@@ -435,7 +431,6 @@ jobs:
         run: |
           source $CONDA/etc/profile.d/conda.sh
           export OCL_ICD_FILENAMES=libintelocl.so
-          export SYCL_ENABLE_HOST_DEVICE=1
           conda activate examples
           conda list
           cd examples/pybind11
@@ -473,7 +468,12 @@ jobs:
             CC=dpcpp CXX=dpcpp LDSHARED="dpcpp -shared" \
                    python setup.py build_ext --inplace || exit 1
             conda deactivate
-            LD_LIBRARY_PATH=${CONDA_PREFIX}/lib python run.py || exit 1
+            if [ -e tests ]
+            then
+               LD_LIBRARY_PATH=${CONDA_PREFIX}/lib python -m pytest tests || exit 1
+            else
+               LD_LIBRARY_PATH=${CONDA_PREFIX}/lib python run.py || exit 1
+            fi
             popd
           done
           cd ../c
@@ -492,7 +492,6 @@ jobs:
           cd examples/python
           source $CONDA/etc/profile.d/conda.sh
           export OCL_ICD_FILENAMES=libintelocl.so
-          export SYCL_ENABLE_HOST_DEVICE=1
           conda activate examples
           for script in $(find . \( -not -name "_*" -and -name "*.py" \))
           do
@@ -598,7 +597,6 @@ jobs:
           conda activate test_dpctl
           # echo "libintelocl.so" | tee /etc/OpenCL/vendors/intel-cpu.icd
           export OCL_ICD_FILENAMES=libintelocl.so
-          export SYCL_ENABLE_HOST_DEVICE=1
           python -c "import dpctl; dpctl.lsplatform()"
           export ARRAY_API_TESTS_MODULE=dpctl.tensor
           cd /home/runner/work/array-api-tests

.github/workflows/generate-coverage.yaml

@@ -85,7 +85,7 @@ jobs:
         shell: bash -l {0}
         run: |
           source /opt/intel/oneapi/setvars.sh
-          SYCL_ENABLE_HOST_DEVICE=1 python scripts/gen_coverage.py
+          python scripts/gen_coverage.py
 
       - name: Install coverall dependencies
         shell: bash -l {0}

.github/workflows/os-llvm-sycl-build.yml

@@ -119,4 +119,4 @@ jobs:
           sycl-ls
           CC=clang CXX=clang++ python setup.py develop -G Ninja
           python -c "import dpctl; dpctl.lsplatform()" || exit 1
-          SYCL_ENABLE_HOST_DEVICE=1 python -m pytest -v dpctl/tests
+          python -m pytest -v dpctl/tests

CHANGELOG.md

@@ -18,6 +18,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Added `dpctl.utils.onetrace_enabled` context manager for targeted trace collection [#903](https://github.com/IntelPython/dpctl/pull/903).
 * Added support for `stream` keyword in `__dlpack__` method, enabling support for sending `usm_ndarray` using mpi4py [#906](https://github.com/IntelPython/dpctl/pull/906).
 * `dpctl.tensor.asarray` can now transition data between incompatible devices, [#951](https://github.com/IntelPython/dpctl/pull/951).
+* Introduced `"syclinterface/dpctl_sycl_types_casters.hpp"` header file with declaration of conversion routines between SYCL type pointers and SyclInterface library opaque pointers [#960](https://github.com/IntelPython/dpctl/pull/960).
+* Added C-API to `dpctl.program.SyclKernel` and `dpctl.program.SyclProgram`. Added type casters for new types to "dpctl4pybind11" and added an example demonstrating its use [#970](https://github.com/IntelPython/dpctl/pull/970).
+* Introduced "dpctl/sycl.pxd" Cython declaration file to streamline use of SYCL functions from Cython, and added an example demonstrating its use [#981](https://github.com/IntelPython/dpctl/pull/981).
+* Added experimental support for sharing data allocated on sub-devices via dlpack [#984](https://github.com/IntelPython/dpctl/pull/984).
+* Added `dpctl.SyclDevice.sub_group_sizes` property to retrieve supported sizes of sub-group by the device [#985](https://github.com/IntelPython/dpctl/pull/985).
 
 ### Changed
 * Improved queue compatibility testing in `dpctl.tensor`'s implementation module [#900](https://github.com/IntelPython/dpctl/pull/900).
@@ -27,13 +32,15 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Refactored `dpctl.tensor`'s implementation module [#941](https://github.com/IntelPython/dpctl/pull/941) to streamline adding new functionality. Streamlined `dpctl::tensor::usm_ndarray` class implementation.
 * Added debugging messaging in case when `DPCTLDynamicLib::getSymbol` encounters errors [#956](https://github.com/IntelPython/dpctl/pull/956).
 * Updated code base according to changes in DPC++ compiler [#952](https://github.com/IntelPython/dpctl/pull/952), [#957](https://github.com/IntelPython/dpctl/pull/957), [#958](https://github.com/IntelPython/dpctl/pull/958).
+* Changed `dpctl` to use pybind11 2.10.1 [#967](https://github.com/IntelPython/dpctl/pull/967).
 
 ### Fixed
 * Improved SyclDevice constructor error message [#893](https://github.com/IntelPython/dpctl/pull/893).
 * Fixed issue gh-890 about `dpctl.tensor.reshape` function [#915](https://github.com/IntelPython/dpctl/pull/915).
 * Fixed unexpected `UnboundLocalError` exception in [#922](https://github.com/IntelPython/dpctl/pull/922).
 * Fixed bugs in `dpctl.tensor.arange` in [#945](https://github.com/IntelPython/dpctl/pull/945).
 * Fixed issue with type inferencing in `dpctl.tensor.asarray` in [#949](https://github.com/IntelPython/dpctl/pull/949).
+* Added missing docstrings for `dpctl.SyclDevice` properties [#964](https://github.com/IntelPython/dpctl/pull/964).
 
 ## [0.13.0] - 07/28/2022
 

MANIFEST.in

@@ -1,4 +1,5 @@
 recursive-include dpctl/include *.h
+recursive-include dpctl/include *.hpp
 include dpctl/include/dpctl4pybind11.hpp
 recursive-include dpctl *.pxd
 recursive-include dpctl *.cmake
@@ -12,6 +13,8 @@ include dpctl/_sycl_event.h
 include dpctl/_sycl_event_api.h
 include dpctl/memory/_memory.h
 include dpctl/memory/_memory_api.h
+include dpctl/program/_program.h
+include dpctl/program/_program_api.h
 include dpctl/tensor/_usmarray.h
 include dpctl/tensor/_usmarray_api.h
 recursive-include dpctl/tensor/include *

conda-recipe/meta.yaml

@@ -26,7 +26,8 @@ requirements:
     run:
         - python
         - {{ pin_compatible('numpy', min_pin='x.x', max_pin='x') }}
-        - dpcpp-cpp-rt >=2022.1
+        - {{ pin_compatible('dpcpp-cpp-rt', min_pin='x.x', max_pin='x') }}  # [py<=39]
+        - dpcpp-cpp-rt  >=2022.2  # [py>39]
 
 test:
     requires:

dpctl/_backend.pxd

@@ -204,6 +204,8 @@ cdef extern from "syclinterface/dpctl_sycl_device_interface.h":
     cdef uint64_t DPCTLDevice_GetGlobalMemCacheSize(const DPCTLSyclDeviceRef DRef)
     cdef _global_mem_cache_type DPCTLDevice_GetGlobalMemCacheType(
         const DPCTLSyclDeviceRef DRef)
+    cdef size_t *DPCTLDevice_GetSubGroupSizes(const DPCTLSyclDeviceRef DRef,
+        size_t *res_len)
 
 
 cdef extern from "syclinterface/dpctl_sycl_device_manager.h":
@@ -263,6 +265,7 @@ cdef extern from "syclinterface/dpctl_sycl_event_interface.h":
 cdef extern from "syclinterface/dpctl_sycl_kernel_interface.h":
     cdef size_t DPCTLKernel_GetNumArgs(const DPCTLSyclKernelRef KRef)
     cdef void DPCTLKernel_Delete(DPCTLSyclKernelRef KRef)
+    cdef DPCTLSyclKernelRef DPCTLKernel_Copy(const DPCTLSyclKernelRef KRef)
     cdef size_t DPCTLKernel_GetWorkGroupSize(const DPCTLSyclKernelRef KRef)
     cdef size_t DPCTLKernel_GetPreferredWorkGroupSizeMultiple(const DPCTLSyclKernelRef KRef)
     cdef size_t DPCTLKernel_GetPrivateMemSize(const DPCTLSyclKernelRef KRef)
@@ -341,6 +344,7 @@ cdef extern from "syclinterface/dpctl_sycl_kernel_bundle_interface.h":
     cdef bool DPCTLKernelBundle_HasKernel(DPCTLSyclKernelBundleRef KBRef,
                                      const char *KernelName)
     cdef void DPCTLKernelBundle_Delete(DPCTLSyclKernelBundleRef KBRef)
+    cdef DPCTLSyclKernelBundleRef DPCTLKernelBundle_Copy(const DPCTLSyclKernelBundleRef KBRef)
 
 
 cdef extern from "syclinterface/dpctl_sycl_queue_interface.h":

dpctl/_device_selection.py

@@ -21,7 +21,7 @@
 from ._sycl_device_factory import get_devices
 
 
-def select_device_with_aspects(required_aspects, excluded_aspects=[]):
+def select_device_with_aspects(required_aspects, excluded_aspects=None):
     """Selects the root :class:`dpctl.SyclDevice` that has the highest
     default selector score among devices that have all aspects in the
     `required_aspects` list, and do not have any aspects in `excluded_aspects`
@@ -41,10 +41,16 @@ def select_device_with_aspects(required_aspects, excluded_aspects=[]):
                 dpctl.select_device_with_aspects(
                     ['usm_shared_allocations'], excluded_aspects=['custom'])
     """
+    if excluded_aspects is None:
+        excluded_aspects = []
     if isinstance(required_aspects, str):
-        required_aspects = [required_aspects]
+        required_aspects = [
+            required_aspects,
+        ]
     if isinstance(excluded_aspects, str):
-        excluded_aspects = [excluded_aspects]
+        excluded_aspects = [
+            excluded_aspects,
+        ]
     seq = collections.abc.Sequence
     input_types_ok = isinstance(required_aspects, seq) and isinstance(
         excluded_aspects, seq
@@ -55,7 +61,7 @@ def select_device_with_aspects(required_aspects, excluded_aspects=[]):
             "e.g. lists, of strings"
         )
     for asp in chain(required_aspects, excluded_aspects):
-        if type(asp) != str:
+        if not isinstance(asp, str):
             raise TypeError("The list objects must be of a string type")
         if not hasattr(SyclDevice, "has_aspect_" + asp):
             raise AttributeError(f"The {asp} aspect is not supported in dpctl")