Merge branch 'feature/advanced-indexing' into take-put-impl

Author: oleksandr-pavlyk

README.md

@@ -3,7 +3,7 @@
 [![pre-commit](https://img.shields.io/badge/pre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white)](https://github.com/pre-commit/pre-commit)
 [![Coverage Status](https://coveralls.io/repos/github/IntelPython/dpctl/badge.svg?branch=master)](https://coveralls.io/github/IntelPython/dpctl?branch=master)
 ![Generate Documentation](https://github.com/IntelPython/dpctl/actions/workflows/generate-docs.yml/badge.svg?branch=master)
-
+[![Join the chat at https://matrix.to/#/#Data-Parallel-Python_community:gitter.im](https://badges.gitter.im/Join%20Chat.svg)](https://app.gitter.im/#/room/#Data-Parallel-Python_community:gitter.im)
 
 <img align="left" src="https://spec.oneapi.io/oneapi-logo-white-scaled.jpg" alt="oneAPI logo" width="75"/>
 

conda-recipe/bld.bat

@@ -6,7 +6,7 @@ set "INCLUDE=%BUILD_PREFIX%\include;%INCLUDE%"
 "%PYTHON%" setup.py clean --all
 set "SKBUILD_ARGS=-G Ninja -- -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icx"
 
-FOR %%V IN (14.0.0 15.0.0 16.0.0) DO @(
+FOR %%V IN (14.0.0 14 15.0.0 15 16.0.0 16) DO @(
   REM set DIR_HINT if directory exists
   IF EXIST "%BUILD_PREFIX%\Library\lib\clang\%%V\" (
      SET "SYCL_INCLUDE_DIR_HINT=%BUILD_PREFIX%\Library\lib\clang\%%V"
@@ -18,11 +18,13 @@ set "FN=Windows-IntelLLVM.cmake"
 
 rem Save the original file, and copy patched file to
 rem fix the issue with IntelLLVM integration with cmake on Windows
-dir "%PLATFORM_DIR%\%FN%"
-copy /Y "%PLATFORM_DIR%\%FN%" .
-if errorlevel 1 exit 1
-copy /Y .github\workflows\Windows-IntelLLVM.cmake "%PLATFORM_DIR%"
-if errorlevel 1 exit 1
+if EXIST "%PLATFORM_DIR%" (
+  dir "%PLATFORM_DIR%\%FN%"
+  copy /Y "%PLATFORM_DIR%\%FN%" .
+  if errorlevel 1 exit 1
+  copy /Y .github\workflows\Windows-IntelLLVM.cmake "%PLATFORM_DIR%"
+  if errorlevel 1 exit 1
+)
 
 if NOT "%WHEELS_OUTPUT_FOLDER%"=="" (
     rem Install and assemble wheel package from the build bits

conda-recipe/build.sh

@@ -30,3 +30,10 @@ else
     # Perform regular install
     ${PYTHON} setup.py install ${SKBUILD_ARGS}
 fi
+
+# need to create this folder so ensure that .dpctl-post-link.sh can work correctly
+mkdir -p $PREFIX/etc/OpenCL/vendors
+echo "dpctl creates symbolic link to system installed /etc/OpenCL/vendors/intel.icd as a work-around." > $PREFIX/etc/OpenCL/vendors/.dpctl_readme
+
+cp $RECIPE_DIR/dpctl-post-link.sh $PREFIX/bin/.dpctl-post-link.sh
+cp $RECIPE_DIR/dpctl-pre-unlink.sh $PREFIX/bin/.dpctl-pre-unlink.sh

conda-recipe/dpctl-post-link.sh

@@ -0,0 +1,9 @@
+#!/usr/bin/bash
+
+systemwide_icd=/etc/OpenCL/vendors/intel.icd
+local_vendors=$PREFIX/etc/OpenCL/vendors/
+icd_fn=$local_vendors/intel-ocl-gpu.icd
+
+if [[ -f $systemwide_icd && -d $local_vendors && ! -f $icd_fn ]]; then
+    ln -s $systemwide_icd $icd_fn
+fi

conda-recipe/dpctl-pre-unlink.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/bash
+
+local_vendors=$PREFIX/etc/OpenCL/vendors/
+icd_fn=$local_vendors/intel-ocl-gpu.icd
+
+if [[ -L $icd_fn ]]; then
+    rm $icd_fn
+fi

conda-recipe/meta.yaml

@@ -12,9 +12,8 @@ build:
 
 requirements:
     build:
-        - {{ compiler('c') }}
         - {{ compiler('cxx') }}
-        - {{ compiler('dpcpp') }}  >=2022.1  # [not osx]
+        - {{ compiler('dpcpp') }}  >=2023.0  # [not osx]
     host:
         - setuptools
         - cmake  >=3.21

dpctl/CMakeLists.txt

@@ -23,7 +23,7 @@ if(WIN32)
     set(CMAKE_CXX_FLAGS_DEBUG
         "${CMAKE_CXX_FLAGS_DEBUG} ${WARNING_FLAGS} -O0 -g1 -DDEBUG"
     )
-    set(DPCTL_LDFLAGS "/link /NXCompat /DynamicBase")
+    set(DPCTL_LDFLAGS "/NXCompat;/DynamicBase")
 elseif(UNIX)
     string(CONCAT WARNING_FLAGS
         "-Wall "
@@ -140,7 +140,8 @@ function(build_dpctl_ext _trgt _src _dest)
         target_compile_options(${_trgt} PRIVATE -fno-sycl-use-footer)
     endif()
     target_link_libraries(${_trgt} DPCTLSyclInterface)
-    target_link_options(${_trgt} PRIVATE "LINKER:${DPCTL_LDFLAGS}")
+    set(_linker_options "LINKER:${DPCTL_LDFLAGS}")
+    target_link_options(${_trgt} PRIVATE ${_linker_options})
     python_extension_module(${_trgt})
     get_filename_component(_name_wle ${_generated_src} NAME_WLE)
     get_filename_component(_generated_src_dir ${_generated_src} DIRECTORY)

dpctl/_backend.pxd

@@ -186,7 +186,14 @@ cdef extern from "syclinterface/dpctl_sycl_device_interface.h":
     cdef uint32_t DPCTLDevice_GetPreferredVectorWidthFloat(const DPCTLSyclDeviceRef DRef)
     cdef uint32_t DPCTLDevice_GetPreferredVectorWidthDouble(const DPCTLSyclDeviceRef DRef)
     cdef uint32_t DPCTLDevice_GetPreferredVectorWidthHalf(const DPCTLSyclDeviceRef DRef)
-    cpdef bool DPCTLDevice_HasAspect(const DPCTLSyclDeviceRef, _aspect_type)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthChar(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthShort(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthInt(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthLong(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthFloat(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthDouble(const DPCTLSyclDeviceRef DRef)
+    cdef uint32_t DPCTLDevice_GetNativeVectorWidthHalf(const DPCTLSyclDeviceRef DRef)
+    cdef bool DPCTLDevice_HasAspect(const DPCTLSyclDeviceRef, _aspect_type)
     cdef uint32_t DPCTLDevice_GetMaxReadImageArgs(const DPCTLSyclDeviceRef DRef)
     cdef uint32_t DPCTLDevice_GetMaxWriteImageArgs(const DPCTLSyclDeviceRef DRef)
     cdef size_t DPCTLDevice_GetImage2dMaxWidth(const DPCTLSyclDeviceRef DRef)

dpctl/_sycl_device.pyx

@@ -54,6 +54,13 @@ from ._backend cimport (  # noqa: E211
     DPCTLDevice_GetMaxWorkItemSizes3d,
     DPCTLDevice_GetMaxWriteImageArgs,
     DPCTLDevice_GetName,
+    DPCTLDevice_GetNativeVectorWidthChar,
+    DPCTLDevice_GetNativeVectorWidthDouble,
+    DPCTLDevice_GetNativeVectorWidthFloat,
+    DPCTLDevice_GetNativeVectorWidthHalf,
+    DPCTLDevice_GetNativeVectorWidthInt,
+    DPCTLDevice_GetNativeVectorWidthLong,
+    DPCTLDevice_GetNativeVectorWidthShort,
     DPCTLDevice_GetParentDevice,
     DPCTLDevice_GetPartitionMaxSubDevices,
     DPCTLDevice_GetPlatform,
@@ -942,6 +949,55 @@ cdef class SyclDevice(_SyclDevice):
         """
         return DPCTLDevice_GetPreferredVectorWidthHalf(self._device_ref)
 
+    @property
+    def native_vector_width_char(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthChar(self._device_ref)
+
+    @property
+    def native_vector_width_short(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthShort(self._device_ref)
+
+    @property
+    def native_vector_width_int(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthInt(self._device_ref)
+
+    @property
+    def native_vector_width_long(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthLong(self._device_ref)
+
+    @property
+    def native_vector_width_float(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthFloat(self._device_ref)
+
+    @property
+    def native_vector_width_double(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthDouble(self._device_ref)
+
+    @property
+    def native_vector_width_half(self):
+        """ Returns the native ISA vector width size for built-in scalar
+        types that can be put into vectors.
+        """
+        return DPCTLDevice_GetNativeVectorWidthHalf(self._device_ref)
+
     @property
     def global_mem_size(self):
         """ Returns the size of global memory on this device in bytes.

dpctl/_sycl_queue_manager.pxd

@@ -17,9 +17,12 @@
 # distutils: language = c++
 # cython: language_level=3
 
+from ._sycl_device cimport SyclDevice
 from ._sycl_queue cimport SyclQueue
 
 
 cpdef SyclQueue get_current_queue()
 cpdef get_current_device_type ()
 cpdef get_current_backend()
+
+cpdef object get_device_cached_queue(object)