Merge pull request #815 from IntelPython/master

merge master to gold

Author: densmirn

.github/workflows/conda-package.yml

@@ -0,0 +1,78 @@
+name: Conda package
+
+on: push
+
+env:
+  PACKAGE_NAME: dpctl
+
+jobs:
+  build:
+    runs-on: ubuntu-20.04
+    steps:
+      - name: Install Intel repository
+        run: |
+          wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2023.PUB
+          sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS-2023.PUB
+          rm GPG-PUB-KEY-INTEL-SW-PRODUCTS-2023.PUB
+          sudo add-apt-repository "deb https://apt.repos.intel.com/oneapi all main"
+          sudo apt-get update
+      - name: Update libstdc++-dev
+        run: |
+          sudo apt remove -y gcc-7 g++-7 gcc-8 g++-8 gcc-10 g++-10
+          sudo apt remove -y libstdc++-10-dev
+          sudo apt autoremove
+          sudo apt install --reinstall -y gcc-9 g++-9 libstdc++-9-dev
+      - name: Install Intel OneAPI
+        run: |
+          sudo apt-get install intel-oneapi-mkl                \
+                               intel-oneapi-mkl-devel          \
+                               intel-oneapi-dpcpp-cpp-compiler
+      - name: Checkout repo
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 0
+
+      - name: Add conda to system path
+        run: echo $CONDA/bin >> $GITHUB_PATH
+      - name: Install conda-build
+        run: conda install conda-build
+      - name: Build conda package
+        run: |
+          . /opt/intel/oneapi/setvars.sh
+          . ./scripts/install_cmake_lin.sh
+          CHANNELS="-c dppy/label/dev -c intel -c defaults --override-channels"
+          VERSIONS="--python 3.8"
+          TEST="--no-test"
+          conda build \
+            $TEST \
+            $VERSIONS \
+            $CHANNELS \
+            conda-recipe
+
+      - name: Upload artifact
+        uses: actions/upload-artifact@v2
+        with:
+          name: dpnp ${{ runner.os }}
+          path: /usr/share/miniconda/conda-bld/linux-64/dpnp-*.tar.bz2
+
+  upload:
+    needs: build
+    if: ${{ github.ref == 'refs/heads/master' }}
+    runs-on: ubuntu-latest
+    steps:
+      - name: Download artifact
+        uses: actions/download-artifact@v2
+        with:
+          name: dpnp ${{ runner.os }}
+
+      - name: Install anaconda-client
+        run: conda install anaconda-client
+      - name: Add conda to system path
+        run: echo $CONDA/bin >> $GITHUB_PATH
+
+      - name: Upload
+        env:
+          ANACONDA_TOKEN: ${{ secrets.ANACONDA_TOKEN }}
+        run: |
+          conda install anaconda-client
+          anaconda --token $ANACONDA_TOKEN upload --user dppy --label dev dpnp-*.tar.bz2

0.build.sh

@@ -17,9 +17,48 @@ python setup.py build_ext --inplace
 
 echo
 echo =========example3==============
-clang++ -g -fPIC dpnp/backend/examples/example3.cpp -Idpnp -Idpnp/backend/include -Ldpnp -Wl,-rpath='$ORIGIN'/dpnp -ldpnp_backend_c -o example3
+dpcpp -g -fPIC dpnp/backend/examples/example3.cpp -Idpnp -Idpnp/backend/include -Ldpnp -Wl,-rpath='$ORIGIN'/dpnp -ldpnp_backend_c -o example3
+# LD_DEBUG=libs,bindings,symbols ./example3
 ./example3
 
-echo
+# gcc --version
+# echo =========LD_LIBRARY_PATH==============
+# echo $LD_LIBRARY_PATH
+
+# echo =========ldd example3==============
+# ldd ./example3
+# echo =========readelf example3==============
+# readelf -d ./example3
+# echo =========ldd dpnp/libdpnp_backend_c.so==============
+# ldd ./dpnp/libdpnp_backend_c.so
+# echo =========readelf dpnp/libdpnp_backend_c.so==============
+# readelf -d ./dpnp/libdpnp_backend_c.so
+
+# echo ========= libstdc++.so ==============
+# ls -l /usr/share/miniconda/envs/dpnp*/lib/libstdc++.so
+# strings /usr/share/miniconda/envs/dpnp*/lib/libstdc++.so | grep GLIBCXX | sort -n
+
+
+# echo
 echo =========example1==============
+# LD_DEBUG=libs,bindings,symbols python examples/example1.py
+# LD_DEBUG=libs python examples/example1.py
 python examples/example1.py
+
+# echo ========= find /opt ==============
+# find /opt -name libstdc++.so*
+# echo ========= find anaconda ==============
+# find /usr/share/miniconda -name libstdc++.so*
+# echo ========= dpkg-query -L libstdc++6 ==============
+# dpkg-query -L libstdc++6
+# echo ========= ls -l /lib/x86_64-linux-gnu/libstdc* ==============
+# ls -l /lib/x86_64-linux-gnu/libstdc*
+
+# gcc --version
+# g++ --version
+# dpcpp --version
+
+# echo ========= APT ==============
+# apt list --installed
+# echo ========= conda ==============
+# conda list

0.env.sh

@@ -9,7 +9,7 @@ export ONEAPI_ROOT=/opt/intel/oneapi
 . ${ONEAPI_ROOT}/compiler/latest/env/vars.sh
 . ${ONEAPI_ROOT}/tbb/latest/env/vars.sh
 
-if true
+if false
 then
     # Temporary use explicit version (arg_verz) due to MKLD-10520
     arg_verz=latest

benchmarks/pytest_benchmark/README.md

@@ -0,0 +1,38 @@
+# dpnp/benchmarks/pytest_benchmark/
+
+## Prerequisites
+* pytest >= 6.1.1
+* pytest-benchmark >= 3.4.1
+
+
+## Running benchmark tests
+```bash
+pytest benchmarks/ --benchmark-json=results.json
+```
+Running tests and saving the current run into `STORAGE`, see [1]
+```bash
+pytest benchmarks/ --benchmark-json=results.json --benchmark-autosave
+```
+
+## Creating `.csv` report
+```bash
+pytest-benchmark compare results.json --csv=results.csv --group-by='name'
+```
+
+## Optional: creating histogram
+Note: make sure that `pytest-benchmark[histogram]` installed 
+```bash
+# example
+pip install pytest-benchmark[histogram]
+pytest -vv benchmarks/ --benchmark-autosave --benchmark-histogram
+pytest-benchmark compare .benchmarks/Linux-CPython-3.7-64bit/* --histogram
+```
+
+## Advanced running example
+```
+pytest benchmarks/ --benchmark-columns='min, max, mean, stddev, median, rounds, iterations' --benchmark-json=results.json --benchmark-autosave
+pytest-benchmark compare results.json --csv=results.csv --group-by='name'
+```
+
+
+[1] https://pytest-benchmark.readthedocs.io/en/latest/usage.html

benchmarks/pytest_benchmark/test_random.py

@@ -0,0 +1,72 @@
+
+# cython: language_level=3
+# -*- coding: utf-8 -*-
+# *****************************************************************************
+# Copyright (c) 2016-2020, Intel Corporation
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+# - Redistributions of source code must retain the above copyright notice,
+#   this list of conditions and the following disclaimer.
+# - Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+# THE POSSIBILITY OF SUCH DAMAGE.
+# *****************************************************************************
+
+import pytest
+
+import dpnp
+import numpy as np
+
+ROUNDS = 30
+ITERATIONS = 4
+
+NNUMBERS = 2**26
+
+
+@pytest.mark.parametrize("function", [dpnp.random.beta, np.random.beta],
+                         ids=["dpnp", "numpy"])
+def test_beta(benchmark, function):
+    result = benchmark.pedantic(target=function, args=(4.0, 5.0, NNUMBERS,),
+                                rounds=ROUNDS, iterations=ITERATIONS)
+
+
+@pytest.mark.parametrize("function", [dpnp.random.exponential, np.random.exponential],
+                         ids=["dpnp", "numpy"])
+def test_exponential(benchmark, function):
+    result = benchmark.pedantic(target=function, args=(4.0, NNUMBERS,),
+                                rounds=ROUNDS, iterations=ITERATIONS)
+
+
+@pytest.mark.parametrize("function", [dpnp.random.gamma, np.random.gamma],
+                         ids=["dpnp", "numpy"])
+def test_gamma(benchmark, function):
+    result = benchmark.pedantic(target=function, args=(2.0, 4.0, NNUMBERS,),
+                                rounds=ROUNDS, iterations=ITERATIONS)
+
+
+@pytest.mark.parametrize("function", [dpnp.random.normal, np.random.normal],
+                         ids=["dpnp", "numpy"])
+def test_normal(benchmark, function):
+    result = benchmark.pedantic(target=function, args=(0.0, 1.0, NNUMBERS,),
+                                rounds=ROUNDS, iterations=ITERATIONS)
+
+
+@pytest.mark.parametrize("function", [dpnp.random.uniform, np.random.uniform],
+                         ids=["dpnp", "numpy"])
+def test_uniform(benchmark, function):
+    result = benchmark.pedantic(target=function, args=(0.0, 1.0, NNUMBERS,),
+                                rounds=ROUNDS, iterations=ITERATIONS)

conda-recipe/build.sh

@@ -16,7 +16,7 @@ fi
 # if MKLROOT is specified (work with custom math library)
 if [ -n "${MKLROOT}" ]; then
     . ${MKLROOT}/env/vars.sh
-    conda remove mkl --force -y || true
+    # conda remove mkl --force -y || true
 fi
 
 # have to activate while SYCL CPU device/driver needs paths

conda-recipe/run_test.sh

@@ -16,7 +16,7 @@ fi
 # if MKLROOT is specified (work with custom math library)
 if [ -n "${MKLROOT}" ]; then
     . ${MKLROOT}/env/vars.sh
-    conda remove mkl --force -y || true
+    # conda remove mkl --force -y || true
 fi
 
 # have to activate while SYCL CPU device/driver needs paths

dpnp/__init__.py

@@ -27,6 +27,16 @@
 import sys
 import warnings
 
+# import os
+# myvar=os.environ
+
+# import pprint
+# pp = pprint.PrettyPrinter(indent=4)
+# pp.pprint(myvar)
+
+# os.system("ldd ./dpnp/libdpnp_backend_c.so")
+# subprocess.call("echo Hello 123", shell=True)
+
 if "numpy" in sys.modules:
     warnings.warn("\nDPNP: Module NumPy found. Please load DPNP module before NumPy.\n")
 

dpnp/backend/CMakeLists.txt

@@ -144,6 +144,7 @@ endif()
 add_definitions(-D_FORTIFY_SOURCE=2 -Wformat -Wformat-security)
 if(NOT WIN32)
   add_compile_options(-fno-delete-null-pointer-checks -fstack-protector-strong -fno-strict-overflow)
+  # add_link_options(-static-libstdc++ -static-libgcc)
   add_link_options(-Wl,-z,noexecstack,-z,relro,-z,now)
 endif()
 
@@ -185,7 +186,9 @@ set(DPNP_SRC
     kernels/dpnp_krnl_statistics.cpp
     src/dpnp_iface_fptr.cpp
     src/memory_sycl.cpp
-    src/queue_sycl.cpp)
+    src/queue_sycl.cpp
+    src/verbose.cpp
+    )
 
 if(DPNP_STATIC_LIB_ENABLE)
   add_library(dpnp_backend_c STATIC ${DPNP_SRC})
@@ -221,7 +224,7 @@ if(UNIX)
   set(DPNP_MATHLIB_DEP_LIBS
       mkl_sycl
       mkl_intel_ilp64
-      mkl_sequential
+      mkl_tbb_thread # mkl_sequential
       mkl_core
       sycl
       OpenCL