Merge master to gold/2021 (#1441)

* Add a copy for strided arrays in dpnp.dot() * update versions * Rework transpose methods to use dpctl.tensor functions (#1389) * Rework transpose methods to call dpctl.tensor functions * Applied review comments & added more tests * Obtain a proper result type on device without fp64 (#1429) * Obtain a proper result type on device without fp64 * remove excess defenitions * Reuse dpctl.tensor.reshape (#1391) * Rework transpose methods to call dpctl.tensor functions * Reuse dpctl.tensor.reshape * added dpnp.shape() and unmuted more tests * fixed compiling issue & unmuted reshaper tests with order param * Resolve merge issues * resolve type mismatch on Win * Use MCG59 engine on GPU device (#1423) * Use MCG59 engine on GPU device * Fix issue for Windows * reduce precision in tests * move w/a before import dpctl * Reuse add(), multiply() and subtract() from dpctl (#1430) * Reuse add(), multiply() and subtract() from dpctl * add in-place support * Reuse dpctl.tensor.sum for dpnp.sum (#1426) * Reuse dpctl.tensor.sun for dpnp.sum * Update tests for dpnp.sum * Fix remarks * Update tests/third_party/cupy/testing/helper.py --------- Co-authored-by: Anton <[email protected]> * Workaround to Klocwork (#1433) * Add inplace support of divide (#1434) * Add dpnp.result_type() support (#1435) * Add dpnp.result_type() support * Update dpnp/dpnp_iface_manipulation.py Co-authored-by: Natalia Polina <[email protected]> --------- Co-authored-by: Natalia Polina <[email protected]> * Implementation of dpnp.mean() (#1431) * Reuse dpctl.tensor.sun for dpnp.sum * Update tests for dpnp.sum * Fix remarks * Implementation of dpnp.mean * Update logic for dpnp.mean function * add normalize_axis_tuple * Additional tests for dpnp.mean * Fix minor remarks * Add inplace support of divide * Use inplace divide only for dpnp.inexact types * Update tests for dpnp.mean * Skip test_sample.py::TestRandint2::test_bound_float1 * Remove unused import * Update dtype check * Update dpnp/dpnp_iface_statistics.py * Return deleted skips --------- Co-authored-by: Anton Volkov <[email protected]> Co-authored-by: Anton <[email protected]> * New implementation of dpnp.outer (#1436) * Add a new implementation of dpnp.outer * Update dpnp.outer implementation --------- Co-authored-by: Anton <[email protected]> * Set dpnp dependency on dpctl>=0.14.3 (#1437) * Implement dpnp.cov() though existing dpnp methods (#1396) * Implement dpnp.cov() though existing dpnp methods * Applied review comments * Clean up the code to get rid of todo * use dpnp.mean() * Added ChangeLog.md (#1439) * Added ChangeLog.md * exclude PR for comprasion ops * Update version to 0.12.0 (#1440) --------- Co-authored-by: vlad-perevezentsev <[email protected]> Co-authored-by: Natalia Polina <[email protected]>
IntelPython · Jun 15, 2023 · 57ce272 · 57ce272
1 parent 67bead1
commit 57ce272
Show file tree

Hide file tree

Showing 51 changed files with 1,737 additions and 1,146 deletions.
diff --git a/.github/workflows/conda-package.yml b/.github/workflows/conda-package.yml
@@ -196,8 +196,6 @@ jobs:
         run: |
           python -m pytest -q -ra --disable-warnings -vv ${{ env.TEST_SCOPE }}
         working-directory: ${{ env.tests-path }}
-        env:
-          SYCL_QUEUE_THREAD_POOL_SIZE: 6
 
   test_windows:
     name: Test ['windows-latest', python='${{ matrix.python }}']
@@ -335,8 +333,6 @@ jobs:
         run: |
           python -m pytest -q -ra --disable-warnings -vv ${{ env.TEST_SCOPE }}
         working-directory: ${{ env.tests-path }}
-        env:
-          SYCL_QUEUE_THREAD_POOL_SIZE: 6
 
   upload:
     name: Upload ['${{ matrix.os }}', python='${{ matrix.python }}']

diff --git a/.github/workflows/generate_coverage.yaml b/.github/workflows/generate_coverage.yaml
@@ -66,7 +66,6 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           COVERALLS_PARALLEL: true
-          SYCL_QUEUE_THREAD_POOL_SIZE: 6
 
   coveralls:
     name: Indicate completion to coveralls.io

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -0,0 +1,55 @@
+# Changelog
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.12.0] - 06/15/2023
+
+### Added
+
+* Implemented `dpnp.broadcast_to` function [#1333](https://github.com/IntelPython/dpnp/pull/1333)
+* Implemented `dpnp.extract` function [#1340](https://github.com/IntelPython/dpnp/pull/1340)
+* Implemented `dpnp.linalg.eigh` function through pybind11 extension of OneMKL call [#1383](https://github.com/IntelPython/dpnp/pull/1383)
+* Implemented `dpnp.mean` function [#1431](https://github.com/IntelPython/dpnp/pull/1431)
+* Added support of bool types in bitwise operations [#1334](https://github.com/IntelPython/dpnp/pull/1334)
+* Added `out` parameter in `dpnp.add` function [#1329](https://github.com/IntelPython/dpnp/pull/1329)
+* Added `out` parameter in `dpnp.multiply` function [#1365](https://github.com/IntelPython/dpnp/pull/1365)
+* Added `out` parameter in `dpnp.sqrt` function [#1332](https://github.com/IntelPython/dpnp/pull/1332)
+* Added `rowvar` parameter in `dpnp.cov` function [#1371](https://github.com/IntelPython/dpnp/pull/1371)
+* Added `nbytes` property to dpnp array [#1359](https://github.com/IntelPython/dpnp/pull/1359)
+* Introduced a new github Action to control code coverage [#1373](https://github.com/IntelPython/dpnp/pull/1373)
+* Added change log [#1439](https://github.com/IntelPython/dpnp/pull/1439)
+
+
+### Changed
+
+* Leveraged `dpctl.tensor` implementation for `dpnp.place` function [#1337](https://github.com/IntelPython/dpnp/pull/1337)
+* Leveraged `dpctl.tensor` implementation for `dpnp.moveaxis` function [#1382](https://github.com/IntelPython/dpnp/pull/1382)
+* Leveraged `dpctl.tensor` implementation for `dpnp.squeeze` function [#1381](https://github.com/IntelPython/dpnp/pull/1381)
+* Leveraged `dpctl.tensor` implementation for `dpnp.where` function [#1380](https://github.com/IntelPython/dpnp/pull/1380)
+* Leveraged `dpctl.tensor` implementation for `dpnp.transpose` function [#1389](https://github.com/IntelPython/dpnp/pull/1389)
+* Leveraged `dpctl.tensor` implementation for `dpnp.reshape` function [#1391](https://github.com/IntelPython/dpnp/pull/1391)
+* Leveraged `dpctl.tensor` implementation for `dpnp.add`, `dpnp.multiply` and `dpnp.subtract` functions [#1430](https://github.com/IntelPython/dpnp/pull/1430)
+* Leveraged `dpctl.tensor` implementation for `dpnp.sum` function [#1426](https://github.com/IntelPython/dpnp/pull/1426)
+* Leveraged `dpctl.tensor` implementation for `dpnp.result_type` function [#1435](https://github.com/IntelPython/dpnp/pull/1435)
+* Reused OneDPL `std::nth_element` function in `dpnp.partition` with 1d array [#1406](https://github.com/IntelPython/dpnp/pull/1406)
+* Transitioned dpnp build system to use scikit-build [#1349](https://github.com/IntelPython/dpnp/pull/1349)
+* Renamed included dpnp_algo_*.pyx files to *.pxi [#1356](https://github.com/IntelPython/dpnp/pull/1356)
+* Implemented support of key as a tuple in `dpnp.__getitem__()` and `dpnp.__setitem__()` functions [#1362](https://github.com/IntelPython/dpnp/pull/1362)
+* Selected dpnp own kernels for elementwise functions instead of OneMKL VM calls on a device without fp64 aspect [#1386](https://github.com/IntelPython/dpnp/pull/1386)
+* Pinned to `sysroot>=2.28` and transitioned to `conda-forge` channel [#1408](https://github.com/IntelPython/dpnp/pull/1408)
+* Redesigned `dpnp.divide` implementation to call `div` from OneMKL for C-contiguous data or to use `dpctl.tensor` library otherwise [#1418](https://github.com/IntelPython/dpnp/pull/1418)
+* Changed an engine used for random generated array on GPU device from MT19937 to MCG59 [#1423](https://github.com/IntelPython/dpnp/pull/1423)
+* Implemented in-place support of `dpnp.divide` [#1434](https://github.com/IntelPython/dpnp/pull/1434)
+* Redesigned `dpnp.outer` implementation through `dpnp.multiply` with broadcasted arrays [#1436](https://github.com/IntelPython/dpnp/pull/1436)
+* Pinned to `dpctl>=0.14.3` as host and run dependencies [#1437](https://github.com/IntelPython/dpnp/pull/1437)
+* Reimplemented `dpnp.cov` through existing dpnp function instead of a separate kernel [#1396](https://github.com/IntelPython/dpnp/pull/1396)
+
+
+### Fixed
+
+* Fixed `dpnp.asarray` function to accept a sequence of dpnp arrays [#1355](https://github.com/IntelPython/dpnp/pull/1355)
+* Fixed crash in `dpnp.sum` with an empty array [#1369](https://github.com/IntelPython/dpnp/pull/1369)
+* Fixed compilation error around `sycl::abs` with DPC++ 2023.2.0 [#1393](https://github.com/IntelPython/dpnp/pull/1393)
+* Fixed Klockwork run and enabled cmake verbose mode for conda build [#1433](https://github.com/IntelPython/dpnp/pull/1433)
diff --git a/conda-recipe/bld.bat b/conda-recipe/bld.bat
@@ -18,7 +18,7 @@ set "DPL_ROOT_HINT=%PREFIX%/Library"
 %PYTHON% -m dpctl --cmakedir > Output
 set /p DPCTL_CMAKE_DIR= < Output
 
-set "SKBUILD_ARGS=-G Ninja -- -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icx"
+set "SKBUILD_ARGS=-G Ninja -- -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icx -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON"
 set "SKBUILD_ARGS=%SKBUILD_ARGS% -DDPCTL_MODULE_PATH:PATH=%DPCTL_CMAKE_DIR% "
 set "SKBUILD_ARGS=%SKBUILD_ARGS% -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON"
 

diff --git a/conda-recipe/build.sh b/conda-recipe/build.sh
@@ -1,5 +1,11 @@
 #!/bin/bash
 
+# Workaround to Klocwork overwriting LD_LIBRARY_PATH that was modified
+# by DPC++ compiler conda packages. Will need to be added to DPC++ compiler
+# activation scripts.
+export LDFLAGS="$LDFLAGS -Wl,-rpath,$PREFIX/lib"
+export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$PREFIX/compiler/lib/intel64_lin:$PREFIX/compiler/lib:$PREFIX/lib"
+
 # Intel LLVM must cooperate with compiler and sysroot from conda
 echo "--gcc-toolchain=${BUILD_PREFIX} --sysroot=${BUILD_PREFIX}/${HOST}/sysroot -target ${HOST}" > icpx_for_conda.cfg
 export ICPXCFG="$(pwd)/icpx_for_conda.cfg"
@@ -10,7 +16,7 @@ export TBB_ROOT_HINT=$PREFIX
 export DPL_ROOT_HINT=$PREFIX
 export MKL_ROOT_HINT=$PREFIX
 SKBUILD_ARGS="-- -DDPCTL_MODULE_PATH=$($PYTHON -m dpctl --cmakedir) "
-SKBUILD_ARGS="${SKBUILD_ARGS} -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icpx"
+SKBUILD_ARGS="${SKBUILD_ARGS} -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icpx -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON"
 SKBUILD_ARGS="${SKBUILD_ARGS} -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON"
 
 # Build wheel package

diff --git a/conda-recipe/meta.yaml b/conda-recipe/meta.yaml
@@ -13,7 +13,7 @@ requirements:
       - cmake >=3.21
       - ninja
       - git
-      - dpctl >=0.14.2
+      - dpctl >=0.14.3
       - mkl-devel-dpcpp {{ environ.get('MKL_VER', '>=2023.1.0') }}
       - onedpl-devel
       - tbb-devel
@@ -25,7 +25,7 @@ requirements:
       - sysroot_linux-64 >=2.28 # [linux]
     run:
       - python
-      - dpctl >=0.14.2
+      - dpctl >=0.14.3
       - {{ pin_compatible('dpcpp-cpp-rt', min_pin='x.x', max_pin='x') }}
       - {{ pin_compatible('mkl-dpcpp', min_pin='x.x', max_pin='x') }}
       - {{ pin_compatible('numpy', min_pin='x.x', max_pin='x') }}

diff --git a/doc/conf.py b/doc/conf.py
@@ -31,9 +31,9 @@
 author = 'Intel'
 
 # The short X.Y version
-version = '0.11'
+version = '0.12'
 # The full version, including alpha/beta/rc tags
-release = '0.11.2dev1'
+release = '0.12.0'
 
 
 # -- General configuration ---------------------------------------------------

diff --git a/dpnp/__init__.py b/dpnp/__init__.py
@@ -27,6 +27,9 @@
 import os
 mypath = os.path.dirname(os.path.realpath(__file__))
 
+# workaround against hanging in OneMKL calls and in DPCTL
+os.environ.setdefault('SYCL_QUEUE_THREAD_POOL_SIZE', '6')
+
 import dpctl
 dpctlpath = os.path.dirname(dpctl.__file__)
 
@@ -41,9 +44,6 @@
         os.add_dll_directory(dpctlpath)
     os.environ["PATH"] = os.pathsep.join([os.getenv("PATH", ""), mypath, dpctlpath])
 
-# workaround against hanging in OneMKL calls
-os.environ.setdefault('SYCL_QUEUE_THREAD_POOL_SIZE', '6')
-
 from dpnp.dpnp_array import dpnp_array as ndarray
 from dpnp.dpnp_flatiter import flatiter as flatiter
 

diff --git a/dpnp/backend/doc/Doxyfile b/dpnp/backend/doc/Doxyfile
@@ -38,7 +38,7 @@ PROJECT_NAME           = "DPNP C++ backend kernel library"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = 0.11.2dev1
+PROJECT_NUMBER         = 0.12.0
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

diff --git a/dpnp/backend/include/dpnp_iface_fptr.hpp b/dpnp/backend/include/dpnp_iface_fptr.hpp
@@ -124,7 +124,6 @@ enum class DPNPFuncName : size_t
     DPNP_FN_COUNT_NONZERO,                /**< Used in numpy.count_nonzero() impl  */
     DPNP_FN_COUNT_NONZERO_EXT,            /**< Used in numpy.count_nonzero() impl, requires extra parameters */
     DPNP_FN_COV,                          /**< Used in numpy.cov() impl  */
-    DPNP_FN_COV_EXT,                      /**< Used in numpy.cov() impl, requires extra parameters */
     DPNP_FN_CROSS,                        /**< Used in numpy.cross() impl  */
     DPNP_FN_CROSS_EXT,                    /**< Used in numpy.cross() impl, requires extra parameters */
     DPNP_FN_CUMPROD,                      /**< Used in numpy.cumprod() impl  */
@@ -218,7 +217,6 @@ enum class DPNPFuncName : size_t
     DPNP_FN_MAXIMUM,                      /**< Used in numpy.maximum() impl  */
     DPNP_FN_MAXIMUM_EXT,                  /**< Used in numpy.maximum() impl , requires extra parameters */
     DPNP_FN_MEAN,                         /**< Used in numpy.mean() impl  */
-    DPNP_FN_MEAN_EXT,                     /**< Used in numpy.mean() impl, requires extra parameters */
     DPNP_FN_MEDIAN,                       /**< Used in numpy.median() impl  */
     DPNP_FN_MEDIAN_EXT,                   /**< Used in numpy.median() impl, requires extra parameters */
     DPNP_FN_MIN,                          /**< Used in numpy.min() impl  */
@@ -360,7 +358,6 @@ enum class DPNPFuncName : size_t
     DPNP_FN_TANH,                         /**< Used in numpy.tanh() impl  */
     DPNP_FN_TANH_EXT,                     /**< Used in numpy.tanh() impl, requires extra parameters */
     DPNP_FN_TRANSPOSE,                    /**< Used in numpy.transpose() impl  */
-    DPNP_FN_TRANSPOSE_EXT,                /**< Used in numpy.transpose() impl, requires extra parameters */
     DPNP_FN_TRACE,                        /**< Used in numpy.trace() impl  */
     DPNP_FN_TRACE_EXT,                    /**< Used in numpy.trace() impl, requires extra parameters */
     DPNP_FN_TRAPZ,                        /**< Used in numpy.trapz() impl  */

diff --git a/dpnp/backend/kernels/dpnp_krnl_manipulation.cpp b/dpnp/backend/kernels/dpnp_krnl_manipulation.cpp
@@ -1,5 +1,5 @@
 //*****************************************************************************
-// Copyright (c) 2016-2020, Intel Corporation
+// Copyright (c) 2016-2023, Intel Corporation
 // All rights reserved.
 //
 // Redistribution and use in source and binary forms, with or without
@@ -211,6 +211,7 @@ void dpnp_elemwise_transpose_c(void* array1_in,
                                                                        size,
                                                                        dep_event_vec_ref);
     DPCTLEvent_WaitAndThrow(event_ref);
+    DPCTLEvent_Delete(event_ref);
 }
 
 template <typename _DataType>
@@ -222,17 +223,6 @@ void (*dpnp_elemwise_transpose_default_c)(void*,
                                           void*,
                                           size_t) = dpnp_elemwise_transpose_c<_DataType>;
 
-template <typename _DataType>
-DPCTLSyclEventRef (*dpnp_elemwise_transpose_ext_c)(DPCTLSyclQueueRef,
-                                                   void*,
-                                                   const shape_elem_type*,
-                                                   const shape_elem_type*,
-                                                   const shape_elem_type*,
-                                                   size_t,
-                                                   void*,
-                                                   size_t,
-                                                   const DPCTLEventVectorRef) = dpnp_elemwise_transpose_c<_DataType>;
-
 void func_map_init_manipulation(func_map_t& fmap)
 {
     fmap[DPNPFuncName::DPNP_FN_REPEAT][eft_INT][eft_INT] = {eft_INT, (void*)dpnp_repeat_default_c<int32_t>};
@@ -253,15 +243,5 @@ void func_map_init_manipulation(func_map_t& fmap)
                                                                (void*)dpnp_elemwise_transpose_default_c<float>};
     fmap[DPNPFuncName::DPNP_FN_TRANSPOSE][eft_DBL][eft_DBL] = {eft_DBL,
                                                                (void*)dpnp_elemwise_transpose_default_c<double>};
-
-    fmap[DPNPFuncName::DPNP_FN_TRANSPOSE_EXT][eft_INT][eft_INT] = {eft_INT,
-                                                                   (void*)dpnp_elemwise_transpose_ext_c<int32_t>};
-    fmap[DPNPFuncName::DPNP_FN_TRANSPOSE_EXT][eft_LNG][eft_LNG] = {eft_LNG,
-                                                                   (void*)dpnp_elemwise_transpose_ext_c<int64_t>};
-    fmap[DPNPFuncName::DPNP_FN_TRANSPOSE_EXT][eft_FLT][eft_FLT] = {eft_FLT,
-                                                                   (void*)dpnp_elemwise_transpose_ext_c<float>};
-    fmap[DPNPFuncName::DPNP_FN_TRANSPOSE_EXT][eft_DBL][eft_DBL] = {eft_DBL,
-                                                                   (void*)dpnp_elemwise_transpose_ext_c<double>};
-
     return;
 }