Merge pull request #1240 from antonwolfy/gold/2021

Merge master into gold/2021

Author: antonwolfy

.github/workflows/build-sphinx.yml

@@ -58,7 +58,7 @@ jobs:
 
       # https://github.com/marketplace/actions/setup-miniconda
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ env.python-ver }}
@@ -106,7 +106,7 @@ jobs:
         if: |
           !github.event.pull_request.head.repo.fork  &&
           (github.ref == 'refs/heads/master' || (startsWith(github.ref, 'refs/heads/release') == true) || github.event_name == 'push' && contains(github.ref, 'refs/tags/'))
-        uses: peaceiris/actions-gh-pages@v3.8.0
+        uses: peaceiris/actions-gh-pages@v3.9.0
         with:
           github_token: ${{ secrets.GITHUB_TOKEN }}
           publish_dir: doc/_build/html/

.github/workflows/conda-package.yml

@@ -10,6 +10,15 @@ env:
   PACKAGE_NAME: dpnp
   MODULE_NAME: dpnp
   CHANNELS: '-c dppy/label/dev -c intel -c defaults --override-channels'
+  TEST_SCOPE: >-
+      test_arraycreation.py
+      test_dparray.py
+      test_fft.py
+      test_linalg.py
+      test_mathematical.py
+      test_random_state.py
+      test_special.py
+      test_usm_type.py
   VER_JSON_NAME: 'version.json'
   VER_SCRIPT1: "import json; f = open('version.json', 'r'); j = json.load(f); f.close(); "
   VER_SCRIPT2: "d = j['dpnp'][0]; print('='.join((d[s] for s in ('version', 'build'))))"
@@ -41,7 +50,7 @@ jobs:
           fetch-depth: 0
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}
@@ -71,7 +80,7 @@ jobs:
         run: conda build --no-test --python ${{ matrix.python }} ${{ env.CHANNELS }} conda-recipe
 
       - name: Upload artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
           path: ${{ env.CONDA_BLD }}${{ env.PACKAGE_NAME }}-*.tar.bz2
@@ -102,7 +111,7 @@ jobs:
           fetch-depth: 0
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}
@@ -134,7 +143,7 @@ jobs:
         run: conda build --no-test --python ${{ matrix.python }} ${{ env.CHANNELS }} conda-recipe
 
       - name: Upload artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
           path: ${{ env.CONDA_BLD }}${{ env.PACKAGE_NAME }}-*.tar.bz2
@@ -165,7 +174,7 @@ jobs:
 
     steps:
       - name: Download artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
           path: ${{ env.pkg-path-in-channel }}
@@ -176,7 +185,7 @@ jobs:
           tar -xvf ${{ env.pkg-path-in-channel }}/${{ env.PACKAGE_NAME }}-*.tar.bz2 -C ${{ env.extracted-pkg-path }}
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}
@@ -235,7 +244,7 @@ jobs:
       # TODO: run the whole scope once the issues on CPU are resolved
       - name: Run tests
         run: |
-          python -m pytest -q -ra --disable-warnings -vv test_arraycreation.py test_dparray.py test_fft.py test_linalg.py test_mathematical.py test_random_state.py test_special.py
+          python -m pytest -q -ra --disable-warnings -vv ${{ env.TEST_SCOPE }}
         env:
           OCL_ICD_FILENAMES: 'libintelocl.so'
         working-directory: ${{ env.tests-path }}
@@ -267,7 +276,7 @@ jobs:
 
     steps:
       - name: Download artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
           path: ${{ env.pkg-path-in-channel }}
@@ -287,7 +296,7 @@ jobs:
           dir ${{ env.extracted-pkg-path }}
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}
@@ -410,14 +419,14 @@ jobs:
       # TODO: run the whole scope once the issues on CPU are resolved
       - name: Run tests
         run: |
-          python -m pytest -q -ra --disable-warnings -vv test_arraycreation.py test_dparray.py test_fft.py test_linalg.py test_mathematical.py test_random_state.py test_special.py
+          python -m pytest -q -ra --disable-warnings -vv ${{ env.TEST_SCOPE }}
         working-directory: ${{ env.tests-path }}
 
   upload_linux:
     needs: test_linux
 
     if: |
-      !github.event.pull_request.head.repo.fork  &&
+      (github.repository == 'IntelPython/dpnp') &&
       (github.ref == 'refs/heads/master' || (startsWith(github.ref, 'refs/heads/release') == true) || github.event_name == 'push' && contains(github.ref, 'refs/tags/'))
 
     runs-on: ubuntu-latest
@@ -432,12 +441,12 @@ jobs:
 
     steps:
       - name: Download artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}
@@ -456,7 +465,7 @@ jobs:
     needs: test_windows
 
     if: |
-      !github.event.pull_request.head.repo.fork  && !github.event.push.repository.fork &&
+      (github.repository == 'IntelPython/dpnp') &&
       (github.ref == 'refs/heads/master' || (startsWith(github.ref, 'refs/heads/release') == true) || github.event_name == 'push' && contains(github.ref, 'refs/tags/'))
 
     runs-on: windows-latest
@@ -470,12 +479,12 @@ jobs:
         python: ['3.8', '3.9']
     steps:
       - name: Download artifact
-        uses: actions/[email protected].0
+        uses: actions/[email protected].1
         with:
           name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
 
       - name: Setup miniconda
-        uses: conda-incubator/setup-miniconda@v2.1.1
+        uses: conda-incubator/setup-miniconda@v2.2.0
         with:
           auto-update-conda: true
           python-version: ${{ matrix.python }}

dpnp/backend/kernels/dpnp_krnl_random.cpp

@@ -973,33 +973,43 @@ DPCTLSyclEventRef dpnp_rng_multinomial_c(DPCTLSyclQueueRef q_ref,
     else
     {
         DPNPC_ptr_adapter<double> p_ptr(q_ref, p_in, p_size, true);
-        const double* p = p_ptr.get_ptr();
-        std::vector<double> p_vec(p, p + p_size);
+        double* p_data = p_ptr.get_ptr();
+
         // size = size
         // `result` is a array for random numbers
-        // `size` is a `result`'s len. `size = n * p.size()`
+        // `size` is a `result`'s len. `size = n * p_size`
         // `n` is a number of random values to be generated.
-        size_t n = size / p_vec.size();
+        size_t n = size / p_size;
 
         size_t is_cpu_queue = dpnp_queue_is_cpu_c();
 
         // math library supports the distribution generation on GPU device with input parameters
         // which follow the condition
         if (is_cpu_queue || (!is_cpu_queue && (p_size >= ((size_t)ntrial * 16)) && (ntrial <= 16)))
         {
-            DPNPC_ptr_adapter<std::int32_t> result_ptr(q_ref, result, size, false, true);
-            std::int32_t* result1 = result_ptr.get_ptr();
-            mkl_rng::multinomial<std::int32_t> distribution(ntrial, p_vec);
+            DPNPC_ptr_adapter<_DataType> result_ptr(q_ref, result, size, true, true);
+            _DataType* result1 = result_ptr.get_ptr();
+
+#if (INTEL_MKL_VERSION < __INTEL_MKL_2023_SWITCHOVER)
+            std::vector<double> p(p_data, p_data + p_size);
+#else
+            auto p = sycl::span<double>{p_data, p_size};
+#endif
+            mkl_rng::multinomial<_DataType> distribution(ntrial, p);
+
             // perform generation
             event_out = mkl_rng::generate(distribution, DPNP_RNG_ENGINE, n, result1);
             event_ref = reinterpret_cast<DPCTLSyclEventRef>(&event_out);
+
+            p_ptr.depends_on(event_out);
+            result_ptr.depends_on(event_out);
         }
         else
         {
-            DPNPC_ptr_adapter<std::int32_t> result_ptr(q_ref, result, size, true, true);
-            std::int32_t* result1 = result_ptr.get_ptr();
+            DPNPC_ptr_adapter<_DataType> result_ptr(q_ref, result, size, true, true);
+            _DataType* result1 = result_ptr.get_ptr();
             int errcode = viRngMultinomial(
-                VSL_RNG_METHOD_MULTINOMIAL_MULTPOISSON, get_rng_stream(), n, result1, ntrial, p_size, p);
+                VSL_RNG_METHOD_MULTINOMIAL_MULTPOISSON, get_rng_stream(), n, result1, ntrial, p_size, p_data);
             if (errcode != VSL_STATUS_OK)
             {
                 throw std::runtime_error("DPNP RNG Error: dpnp_rng_multinomial_c() failed.");
@@ -1023,6 +1033,7 @@ void dpnp_rng_multinomial_c(
                                                                     size,
                                                                     dep_event_vec_ref);
     DPCTLEvent_WaitAndThrow(event_ref);
+    DPCTLEvent_Delete(event_ref);
 }
 
 template <typename _DataType>
@@ -1065,24 +1076,32 @@ DPCTLSyclEventRef dpnp_rng_multivariate_normal_c(DPCTLSyclQueueRef q_ref,
     sycl::queue q = *(reinterpret_cast<sycl::queue*>(q_ref));
 
     DPNPC_ptr_adapter<double> mean_ptr(q_ref, mean_in, mean_size, true);
-    const double* mean = mean_ptr.get_ptr();
+    double* mean_data = mean_ptr.get_ptr();
     DPNPC_ptr_adapter<double> cov_ptr(q_ref, cov_in, cov_size, true);
-    const double* cov = cov_ptr.get_ptr();
+    double* cov_data = cov_ptr.get_ptr();
 
-    _DataType* result1 = reinterpret_cast<_DataType*>(result);
+    _DataType* result1 = static_cast<_DataType *>(result);
 
-    std::vector<double> mean_vec(mean, mean + mean_size);
-    std::vector<double> cov_vec(cov, cov + cov_size);
+#if (INTEL_MKL_VERSION < __INTEL_MKL_2023_SWITCHOVER)
+    std::vector<double> mean(mean_data, mean_data + mean_size);
+    std::vector<double> cov(cov_data, cov_data + cov_size);
+#else
+    auto mean = sycl::span<double>{mean_data, mean_size};
+    auto cov = sycl::span<double>{cov_data, cov_size};
+#endif
 
     // `result` is a array for random numbers
     // `size` is a `result`'s len.
     // `size1` is a number of random values to be generated for each dimension.
     size_t size1 = size / dimen;
 
-    mkl_rng::gaussian_mv<_DataType> distribution(dimen, mean_vec, cov_vec);
+    mkl_rng::gaussian_mv<_DataType> distribution(dimen, mean, cov);
     auto event_out = mkl_rng::generate(distribution, DPNP_RNG_ENGINE, size1, result1);
     event_ref = reinterpret_cast<DPCTLSyclEventRef>(&event_out);
 
+    mean_ptr.depends_on(event_out);
+    cov_ptr.depends_on(event_out);
+
     return DPCTLEvent_Copy(event_ref);
 }
 
@@ -1107,6 +1126,7 @@ void dpnp_rng_multivariate_normal_c(void* result,
                                                                             size,
                                                                             dep_event_vec_ref);
     DPCTLEvent_WaitAndThrow(event_ref);
+    DPCTLEvent_Delete(event_ref);
 }
 
 template <typename _DataType>

dpnp/backend/src/dpnp_utils.hpp

@@ -49,6 +49,13 @@
 #define __SYCL_COMPILER_2023_SWITCHOVER 20221102L
 #endif
 
+/**
+ * Version of Intel MKL at which transition to OneMKL release 2023.0.0 occurs.
+ */
+#ifndef __INTEL_MKL_2023_SWITCHOVER
+#define __INTEL_MKL_2023_SWITCHOVER 20230000
+#endif
+
 /**
  * @defgroup BACKEND_UTILS Backend C++ library utilities
  * @{

dpnp/dpnp_utils/dpnp_algo_utils.pyx

@@ -532,22 +532,7 @@ cdef tuple get_common_usm_allocation(dpnp_descriptor x1, dpnp_descriptor x2):
     array1_obj = x1.get_array()
     array2_obj = x2.get_array()
 
-    def get_usm_type(usm_types):
-        if not isinstance(usm_types, (list, tuple)):
-            raise TypeError(
-                "Expected a list or a tuple, got {}".format(type(usm_types))
-            )
-        if len(usm_types) == 0:
-            return None
-        elif len(usm_types) == 1:
-            return usm_types[0]
-        for usm_type1, usm_type2 in zip(usm_types, usm_types[1:]):
-            if usm_type1 != usm_type2:
-                return None
-        return usm_types[0]
-
-    # TODO: use similar function from dpctl.utils instead of get_usm_type
-    common_usm_type = get_usm_type((array1_obj.usm_type, array2_obj.usm_type))
+    common_usm_type = dpctl.utils.get_coerced_usm_type((array1_obj.usm_type, array2_obj.usm_type))
     if common_usm_type is None:
         raise ValueError(
             "could not recognize common USM type for inputs of USM types {} and {}"

tests/skipped_tests.tbl

@@ -199,9 +199,12 @@ tests/test_linalg.py::test_svd[(5,3)-float64]
 tests/test_linalg.py::test_svd[(16,16)-float64]
 
 tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[3.5-array1]
+
+tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray([[i, i] for i in x])]
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: (dpnp.asarray([(i, i) for i in x], [("a", int), ("b", int)]).view(dpnp.recarray))]
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray([(i, i) for i in x], [("a", object), ("b", dpnp.int32)])]]
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.asarray(x).astype(dpnp.int8)]
+
 tests/test_sort.py::test_partition[[[1, 0], [3, 0]]-float32-1]
 tests/third_party/cupy/binary_tests/test_elementwise.py::TestElementwise::test_bitwise_and
 tests/third_party/cupy/binary_tests/test_elementwise.py::TestElementwise::test_bitwise_or

tests/skipped_tests_gpu.tbl

@@ -1,5 +1,3 @@
-tests/test_random.py::TestDistributionsMultinomial::test_extreme_value
-tests/test_random.py::TestDistributionsMultinomial::test_seed1
 tests/test_random.py::TestDistributionsMultivariateNormal::test_moments
 tests/test_random.py::TestDistributionsMultivariateNormal::test_output_shape_check
 tests/test_random.py::TestDistributionsMultivariateNormal::test_seed
@@ -219,9 +217,7 @@ tests/test_mathematical.py::TestGradient::test_gradient_y1[array2]
 tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array0]
 tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array1]
 tests/test_mathematical.py::TestGradient::test_gradient_y1_dx[2-array2]
-tests/test_random.py::TestDistributionsMultinomial::test_check_sum
-tests/test_random.py::TestDistributionsMultinomial::test_moments
-tests/test_random.py::TestDistributionsMultinomial::test_seed
+
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.astype(dpnp.asarray(x), dpnp.int8)]
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.astype(dpnp.asarray(x), object)]
 tests/test_random.py::TestPermutationsTestShuffle::test_shuffle1[lambda x: dpnp.vstack([x, x]).T]

tests/test_usm_type.py

@@ -0,0 +1,36 @@
+import pytest
+
+import dpnp as dp
+
+import dpctl.utils as du
+
+list_of_usm_types = [
+    "device",
+    "shared",
+    "host"
+]
+
+
+@pytest.mark.parametrize("usm_type", list_of_usm_types, ids=list_of_usm_types)
+def test_coerced_usm_types_sum(usm_type):
+    x = dp.arange(10, usm_type = "device")
+    y = dp.arange(10, usm_type = usm_type)
+
+    z = x + y
+    
+    assert z.usm_type == x.usm_type
+    assert z.usm_type == "device"
+    assert y.usm_type == usm_type
+
+
+@pytest.mark.parametrize("usm_type_x", list_of_usm_types, ids=list_of_usm_types)
+@pytest.mark.parametrize("usm_type_y", list_of_usm_types, ids=list_of_usm_types)
+def test_coerced_usm_types_mul(usm_type_x, usm_type_y):
+    x = dp.arange(10, usm_type = usm_type_x)
+    y = dp.arange(10, usm_type = usm_type_y)
+
+    z = x * y
+    
+    assert x.usm_type == usm_type_x
+    assert y.usm_type == usm_type_y
+    assert z.usm_type == du.get_coerced_usm_type([usm_type_x, usm_type_y])