Python array API conformity workflow added#881

Closed

ndgrigorian wants to merge 102 commits intoIntelPython:masterfrom

ndgrigorian:array-api-conformity

Commits on Aug 12, 2022

Python array API conformity workflow added
ndgrigorian
committed

Commits on Aug 16, 2022

Fixed whitespace
ndgrigorian
authored

Commits on Aug 22, 2022

Merge branch 'IntelPython:master' into array-api-conformity
ndgrigorian
authored
Python array API conformity workflow added
ndgrigorian
committed
dpctl.tensor.asarray must check numpy array data-type

oleksandr-pavlyk
authored and
ndgrigorian
committed
Altered the logic of finding the most recent tag to download from

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added a test to check validation for supported dtype

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed test to pass on Iris Xe

oleksandr-pavlyk
authored and
ndgrigorian
committed
Removing stray print

oleksandr-pavlyk
authored and
ndgrigorian
committed
Correcting exception text

oleksandr-pavlyk
authored and
ndgrigorian
committed
Wrap SyclQueue constructor in try/catch

oleksandr-pavlyk
authored and
ndgrigorian
committed
dpctl.tensor.from_numpy should not try creating USM ndarray for 64-bit fp on HW that has no such support

oleksandr-pavlyk
authored and
ndgrigorian
committed
Replaced constructor with dtype 'd' with call to ones to dynamically figure out the data type appropriate for the device

oleksandr-pavlyk
authored and
ndgrigorian
committed
Update array-api-conformity.yml
ndgrigorian
committed
Merge branch 'IntelPython:master' into array-api-conformity-test
ndgrigorian
authored

Commits on Aug 23, 2022

Commits on Aug 31, 2022

Merge branch 'IntelPython:master' into array-api-conformity
ndgrigorian
authored

Commits on Sep 8, 2022

Adjustments to account for changed include location in main trunk of DPC++

oleksandr-pavlyk
authored and
ndgrigorian
committed
Make sure to catch SYCL exception around event::wait calls

oleksandr-pavlyk
authored and
ndgrigorian
committed
Do not submit kernels if HW does not support double precision

oleksandr-pavlyk
authored and
ndgrigorian
committed
Be explicit about allocation queue in tests

oleksandr-pavlyk
authored and
ndgrigorian
committed
Using quotes for dpctl/CMakeLists.txt per Anton's feedback

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed stray use of deprecated SYCL-2020 quantifier in dpctl_c_api tests

oleksandr-pavlyk
authored and
ndgrigorian
committed
Changed test from copying doubles to copying int64 to run on Iris Xe

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added testing logic to skip test cases resulting in JIT-ting kernel for double precision type is the HW does not support it

oleksandr-pavlyk
authored and
ndgrigorian
committed
Getter functions for usm_ndarray are marked const per gh-852

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed typos in docstrings found by Doxygen

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fix issues tripping documentation build

oleksandr-pavlyk
authored and
ndgrigorian
committed
expanded changelog

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added license text to IntelDPCPPConfig.cmake to align it with what's is oneAPI DPC++

oleksandr-pavlyk
authored and
ndgrigorian
committed
Adding dpctl.tensor.concat feature and tests.

npolina4
authored and
ndgrigorian
committed
Use numpy.promote_types to allow concatenation of different dtypes

oleksandr-pavlyk
authored and
ndgrigorian
committed
expanded changelog

npolina4
authored and
ndgrigorian
committed
Fixed filter-selector-string output by lspltform and print_device_info

oleksandr-pavlyk
authored and
ndgrigorian
committed
Improved exceptions msg

oleksandr-pavlyk
authored and
ndgrigorian
committed
Closes gh-870

oleksandr-pavlyk
authored and
ndgrigorian
committed
Adds tests for gh-870

oleksandr-pavlyk
authored and
ndgrigorian
committed
Noted the change in CHANGELOG.md

oleksandr-pavlyk
authored and
ndgrigorian
committed
Adds dpctl.tensor.concat test for gh-870

npolina4
authored and
ndgrigorian
committed
Added dpctl.tensor.stack feature and tests

npolina4
authored and
ndgrigorian
committed
Renames an internal function

npolina4
authored and
ndgrigorian
committed
Noted the change in CHANGELOG.md

npolina4
authored and
ndgrigorian
committed
Updated changelog in preparation for release of 0.13.0

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added test to check that assigned of different types of the same size work without errors

oleksandr-pavlyk
authored and
ndgrigorian
committed
typenum_to_src_id now recognized shadowed types, i.e. intc on Windows

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed compiler warning

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed arange(3.) to not produce float64 array on HW with fp64 support

oleksandr-pavlyk
authored and
ndgrigorian
committed
asarray on numpy arrays should take device capabilities into account in decision in data-type of the USM array

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added array-API dpctl.tensor.linspace

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added changelog entry fo #875

oleksandr-pavlyk
authored and
ndgrigorian
committed
pybind11 v2.9.2 -> v2.10.0
ndgrigorian
committed
Simplified impl of tensor::usm_ndarray and memory::usm_memory classes

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fix build break with open-source intel/llvm DPC++

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added max_work_item_sizes1d, 2d, 3d properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added tests for new device properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added tests for deprecated property:

oleksandr-pavlyk
authored and
ndgrigorian
committed
linspace_affine should not use double precision type in kernels is HW does not support it

oleksandr-pavlyk
authored and
ndgrigorian
committed
test_linspace should not try double precision if HW does not support it

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fix for copy of double precision NumPy array to device w/o HW support for DP

oleksandr-pavlyk
authored and
ndgrigorian
committed
Tests should not use double precision arrays where it is not essential for testing

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added example of C extension using DPCTLSyclInterface

oleksandr-pavlyk
authored and
ndgrigorian
committed
Use [[deprecated]] attributed supported in both gcc/g++ and icx/icpx

oleksandr-pavlyk
authored and
ndgrigorian
committed
Replaced print with use of warnings.warn

oleksandr-pavlyk
authored and
ndgrigorian
committed
Add building of native Python extension written in C to test_linux_examples

oleksandr-pavlyk
authored and
ndgrigorian
committed
Removed use of deprecated methods

oleksandr-pavlyk
authored and
ndgrigorian
committed
Removed deprecated symbols, set library_dirs for setuptools builds

oleksandr-pavlyk
authored and
ndgrigorian
committed
dpctl.tensor.asarray must check numpy array data-type

oleksandr-pavlyk
authored and
ndgrigorian
committed
Altered the logic of finding the most recent tag to download from

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added a test to check validation for supported dtype

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed test to pass on Iris Xe

oleksandr-pavlyk
authored and
ndgrigorian
committed
Removing stray print

oleksandr-pavlyk
authored and
ndgrigorian
committed
Correcting exception text

oleksandr-pavlyk
authored and
ndgrigorian
committed
Wrap SyclQueue constructor in try/catch

oleksandr-pavlyk
authored and
ndgrigorian
committed
dpctl.tensor.from_numpy should not try creating USM ndarray for 64-bit fp on HW that has no such support

oleksandr-pavlyk
authored and
ndgrigorian
committed
Replaced constructor with dtype 'd' with call to ones to dynamically figure out the data type appropriate for the device

oleksandr-pavlyk
authored and
ndgrigorian
committed
Adjust script setting hint dir variable to work with sycl bundles using different clang versions

oleksandr-pavlyk
authored and
ndgrigorian
committed
static_cast to void pointer from char pointer in calls to memcpy to appease klocwork scanner

oleksandr-pavlyk
authored and
ndgrigorian
committed
Use consistent order of channels during build and test phases

oleksandr-pavlyk
authored and
ndgrigorian
committed
Import error_already_set into namespace

oleksandr-pavlyk
authored and
ndgrigorian
committed
Improved exception message when SyclDevice could not created from invalid filter string

oleksandr-pavlyk
authored and
ndgrigorian
committed
Use [[deprecated("msg")]] annotation only for C++ builds

oleksandr-pavlyk
authored and
ndgrigorian
committed
[[deprecated]] is part of C23 standard, it is supported in gcc version 10 and older

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added 3 new device attributes per gh-886

oleksandr-pavlyk
authored and
ndgrigorian
committed
Expanded checks for attributes of SyclDevice to include all properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added 3 more SyclDevice properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added all but one kernel_device_specific properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Private mem size may be 0 on CPU devices

oleksandr-pavlyk
authored and
ndgrigorian
committed
Exported kernel device-specific properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Improved docstring for DPCTLSyclKernelBundleRef

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added tests for kernel_device_specific properties

oleksandr-pavlyk
authored and
ndgrigorian
committed
Added property dpctl.program.SyclKernel.num_args

oleksandr-pavlyk
authored and
ndgrigorian
committed
Fixed test for sycl_kernel, added test for exercising equal testing for SyclDevice and object of another type

oleksandr-pavlyk
authored and
ndgrigorian
committed
Adrressed PR review feedback

oleksandr-pavlyk
authored and
ndgrigorian
committed
Array API tests reworked
ndgrigorian
committed
Merge branch 'array-api-conformity' of https://github.com/ndgrigorian/dpctl into array-api-conformity
ndgrigorian
committed