pyopencl 2023.1.4__tar.gz → 2024.2__tar.gz

{pyopencl-2023.1.4 → pyopencl-2024.2}/CITATION.cff

@@ -71,3 +71,4 @@ version: 2022.1.3
 doi: 10.5281/zenodo.6533956
 date-released: 2022-03-10
 url: "https://github.com/inducer/pyopencl"
+license: MIT

{pyopencl-2023.1.4 → pyopencl-2024.2}/LICENSE

@@ -78,26 +78,3 @@ PyOpenCL includes parts of the Random123 suite of random number generators:
     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.
-
-PyOpenCL includes the RANLUXCL random number generator:
-
-    Copyright (c) 2011 Ivar Ursin Nikolaisen
-
-    Permission is hereby granted, free of charge, to any person obtaining a copy of this
-    software and associated documentation files (the "Software"), to deal in the Software
-    without restriction, including without limitation the rights to use, copy, modify,
-    merge, publish, distribute, sublicense, and/or sell copies of the Software, and to
-    permit persons to whom the Software is furnished to do so, subject to the following
-    conditions:
-
-    The above copyright notice and this permission notice shall be included in all copies
-    or substantial portions of the Software.
-
-    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
-    INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
-    PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
-    HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
-    CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
-    OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-

{pyopencl-2023.1.4 → pyopencl-2024.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pyopencl
-Version: 2023.1.4
+Version: 2024.2
 Summary: Python wrapper for OpenCL
 Home-page: http://mathema.tician.de/software/pyopencl
 Author: Andreas Kloeckner
@@ -22,8 +22,9 @@ Classifier: Topic :: Scientific/Engineering :: Physics
 Requires-Python: ~=3.8
 License-File: LICENSE
 Requires-Dist: numpy
-Requires-Dist: pytools>=2021.2.7
+Requires-Dist: pytools>=2022.1.13
 Requires-Dist: platformdirs>=2.2.0
+Requires-Dist: importlib-resources; python_version < "3.9"
 Provides-Extra: pocl
 Requires-Dist: pocl_binary_distribution>=1.2; extra == "pocl"
 Provides-Extra: oclgrind
@@ -35,19 +36,21 @@ Requires-Dist: Mako; extra == "test"
 PyOpenCL: Pythonic Access to OpenCL, with Arrays and Algorithms
 ===============================================================
 
-.. image:: https://gitlab.tiker.net/inducer/pyopencl/badges/main/pipeline.svg
+.. |badge-gitlab-ci| image:: https://gitlab.tiker.net/inducer/pyopencl/badges/main/pipeline.svg
     :alt: Gitlab Build Status
     :target: https://gitlab.tiker.net/inducer/pyopencl/commits/main
-.. image:: https://github.com/inducer/pyopencl/workflows/CI/badge.svg?branch=main&event=push
+.. |badge-github-ci| image:: https://github.com/inducer/pyopencl/workflows/CI/badge.svg?branch=main&event=push
     :alt: Github Build Status
     :target: https://github.com/inducer/pyopencl/actions?query=branch%3Amain+workflow%3ACI+event%3Apush
-.. image:: https://badge.fury.io/py/pyopencl.svg
+.. |badge-pypi| image:: https://badge.fury.io/py/pyopencl.svg
     :alt: Python Package Index Release Page
     :target: https://pypi.org/project/pyopencl/
-.. image:: https://zenodo.org/badge/1575307.svg
+.. |badge-zenodo| image:: https://zenodo.org/badge/1575307.svg
     :alt: Zenodo DOI for latest release
     :target: https://zenodo.org/badge/latestdoi/1575307
 
+|badge-gitlab-ci| |badge-github-ci| |badge-pypi| |badge-zenodo|
+
 PyOpenCL lets you access GPUs and other massively parallel compute
 devices from Python. It tries to offer computing goodness in the
 spirit of its sister project `PyCUDA <https://mathema.tician.de/software/pycuda>`__:
@@ -84,8 +87,7 @@ can be found in the `documentation <https://documen.tician.de/pyopencl/>`__.
 What you'll need if you do *not* want to use the convenient instructions above and
 instead build from source:
 
-* gcc/g++ new enough to be compatible with pybind11
-  (see their `FAQ <https://pybind11.readthedocs.io/en/stable/faq.html>`__)
+* g++/clang new enough to be compatible with nanobind (specifically, full support of C++17 is needed)
 * `numpy <https://numpy.org>`__, and
 * an OpenCL implementation. (See this `howto <https://wiki.tiker.net/OpenCLHowTo>`__
   for how to get one.)
@@ -95,11 +97,9 @@ Links
 
 * `Documentation <https://documen.tician.de/pyopencl>`__
   (read how things work)
+* `Python package index <https://pypi.python.org/pypi/pyopencl>`__
+  (download releases, including binary wheels for Linux, macOS, Windows)
 * `Conda Forge <https://anaconda.org/conda-forge/pyopencl>`__
   (download binary packages for Linux, macOS, Windows)
-* `Python package index <https://pypi.python.org/pypi/pyopencl>`__
-  (download releases)
-* `C. Gohlke's Windows binaries <https://www.lfd.uci.edu/~gohlke/pythonlibs/#pyopencl>`__
-  (download Windows binaries)
 * `Github <https://github.com/inducer/pyopencl>`__
   (get latest source code, file bugs)

{pyopencl-2023.1.4 → pyopencl-2024.2}/README.rst

@@ -1,19 +1,21 @@
 PyOpenCL: Pythonic Access to OpenCL, with Arrays and Algorithms
 ===============================================================
 
-.. image:: https://gitlab.tiker.net/inducer/pyopencl/badges/main/pipeline.svg
+.. |badge-gitlab-ci| image:: https://gitlab.tiker.net/inducer/pyopencl/badges/main/pipeline.svg
     :alt: Gitlab Build Status
     :target: https://gitlab.tiker.net/inducer/pyopencl/commits/main
-.. image:: https://github.com/inducer/pyopencl/workflows/CI/badge.svg?branch=main&event=push
+.. |badge-github-ci| image:: https://github.com/inducer/pyopencl/workflows/CI/badge.svg?branch=main&event=push
     :alt: Github Build Status
     :target: https://github.com/inducer/pyopencl/actions?query=branch%3Amain+workflow%3ACI+event%3Apush
-.. image:: https://badge.fury.io/py/pyopencl.svg
+.. |badge-pypi| image:: https://badge.fury.io/py/pyopencl.svg
     :alt: Python Package Index Release Page
     :target: https://pypi.org/project/pyopencl/
-.. image:: https://zenodo.org/badge/1575307.svg
+.. |badge-zenodo| image:: https://zenodo.org/badge/1575307.svg
     :alt: Zenodo DOI for latest release
     :target: https://zenodo.org/badge/latestdoi/1575307
 
+|badge-gitlab-ci| |badge-github-ci| |badge-pypi| |badge-zenodo|
+
 PyOpenCL lets you access GPUs and other massively parallel compute
 devices from Python. It tries to offer computing goodness in the
 spirit of its sister project `PyCUDA <https://mathema.tician.de/software/pycuda>`__:
@@ -50,8 +52,7 @@ can be found in the `documentation <https://documen.tician.de/pyopencl/>`__.
 What you'll need if you do *not* want to use the convenient instructions above and
 instead build from source:
 
-* gcc/g++ new enough to be compatible with pybind11
-  (see their `FAQ <https://pybind11.readthedocs.io/en/stable/faq.html>`__)
+* g++/clang new enough to be compatible with nanobind (specifically, full support of C++17 is needed)
 * `numpy <https://numpy.org>`__, and
 * an OpenCL implementation. (See this `howto <https://wiki.tiker.net/OpenCLHowTo>`__
   for how to get one.)
@@ -61,11 +62,9 @@ Links
 
 * `Documentation <https://documen.tician.de/pyopencl>`__
   (read how things work)
+* `Python package index <https://pypi.python.org/pypi/pyopencl>`__
+  (download releases, including binary wheels for Linux, macOS, Windows)
 * `Conda Forge <https://anaconda.org/conda-forge/pyopencl>`__
   (download binary packages for Linux, macOS, Windows)
-* `Python package index <https://pypi.python.org/pypi/pyopencl>`__
-  (download releases)
-* `C. Gohlke's Windows binaries <https://www.lfd.uci.edu/~gohlke/pythonlibs/#pyopencl>`__
-  (download Windows binaries)
 * `Github <https://github.com/inducer/pyopencl>`__
   (get latest source code, file bugs)

{pyopencl-2023.1.4 → pyopencl-2024.2}/aksetup_helper.py

@@ -1,9 +1,11 @@
 import os
 import sys
+
+
 try:
     from setuptools import Extension
-    from setuptools.command.build_ext import (  # noqa: N812
-            build_ext as BaseBuildExtCommand)
+    from setuptools.command.build_ext import \
+        build_ext as BaseBuildExtCommand  # noqa: N812
 
 except ImportError:
     class Extension:
@@ -43,14 +45,21 @@ def setup(*args, **kwargs):
 
 
 def get_numpy_incpath():
-    from os.path import join, dirname
     from importlib.util import find_spec
+    from os.path import dirname, exists, join
     origin = find_spec("numpy").origin
     if origin is None:
         raise RuntimeError("origin of numpy package not found")
 
     pathname = dirname(origin)
-    return join(pathname, "core", "include")
+    for p in [
+        join(pathname, "_core", "include"),  # numpy 2 onward
+        join(pathname, "core", "include"),  # numpy prior to 2
+    ]:
+        if exists(join(p, "numpy", "arrayobject.h")):
+            return p
+
+    raise RuntimeError("no valid path for numpy found")
 
 
 class NumpyExtension(Extension):
@@ -762,7 +771,7 @@ def substitute(substitutions, fname):
     with open(fname, "w") as outf:
         outf.write("".join(new_lines))
 
-    from os import stat, chmod
+    from os import chmod, stat
     infile_stat_res = stat(fname_in)
     chmod(fname, infile_stat_res.st_mode)
 
@@ -771,7 +780,7 @@ def substitute(substitutions, fname):
 
 def _run_git_command(cmd):
     git_error = None
-    from subprocess import Popen, PIPE
+    from subprocess import PIPE, Popen
     stdout = None
     try:
         popen = Popen(["git"] + cmd, stdout=PIPE)

{pyopencl-2023.1.4 → pyopencl-2024.2}/configure.py

@@ -1,4 +1,6 @@
 #! /usr/bin/env python3
 
 from aksetup_helper import configure_frontend
+
+
 configure_frontend()

{pyopencl-2023.1.4 → pyopencl-2024.2}/contrib/cldis.py

@@ -28,12 +28,12 @@ A script to compile and print the native code for a OpenCL kernel.
 Usage: python cldis.py prog.cl <ptx/sass/asm> <build options>
 """
 
+import glob
 import os
-import tempfile
-import subprocess
 import re
+import subprocess
 import sys
-import glob
+import tempfile
 
 
 def main(ctx, tmp_dir, cl_str, output=None, build_options=[]):
@@ -44,7 +44,7 @@ def main(ctx, tmp_dir, cl_str, output=None, build_options=[]):
     elif platform.name == "Portable Computing Language":
         if device.name.startswith("NVIDIA"):
             supported_outputs = ["ptx", "sass"]
-        elif device.name.startswith("pthread"):
+        elif device.name.startswith("pthread") or device.name.startswith("cpu")
             supported_outputs = ["asm"]
         else:
             raise NotImplementedError(f"Unknown pocl device '{device.name}'")

{pyopencl-2023.1.4 → pyopencl-2024.2}/contrib/fortran-to-opencl/translate.py

@@ -20,18 +20,17 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE.
 """
 
-import cgen
-import numpy as np
 import re
-from pymbolic.parser import Parser as ExpressionParserBase
-from pymbolic.mapper import CombineMapper
-import pymbolic.primitives as p
-from pymbolic.mapper.c_code import CCodeMapper as CCodeMapperBase
 from sys import intern
-
 from warnings import warn
 
+import cgen
+import numpy as np
+import pymbolic.primitives as p
 import pytools.lex
+from pymbolic.mapper import CombineMapper
+from pymbolic.mapper.c_code import CCodeMapper as CCodeMapperBase
+from pymbolic.parser import Parser as ExpressionParserBase
 
 
 class TranslatorWarning(UserWarning):
@@ -154,8 +153,7 @@ class FortranExpressionParser(ExpressionParserBase):
     def parse_terminal(self, pstate):
         scope = self.tree_walker.scope_stack[-1]
 
-        from pymbolic.parser import (
-            _identifier, _openpar, _closepar, _float)
+        from pymbolic.parser import _closepar, _float, _identifier, _openpar
 
         next_tag = pstate.next_tag()
         if next_tag is _float:
@@ -221,8 +219,8 @@ class FortranExpressionParser(ExpressionParserBase):
             }
 
     def parse_prefix(self, pstate, min_precedence=0):
-        from pymbolic.parser import _PREC_UNARY
         import pymbolic.primitives as primitives
+        from pymbolic.parser import _PREC_UNARY
 
         pstate.expect_not_end()
 
@@ -235,10 +233,9 @@ class FortranExpressionParser(ExpressionParserBase):
 
     def parse_postfix(self, pstate, min_precedence, left_exp):
         from pymbolic.parser import (
-                _PREC_CALL, _PREC_COMPARISON, _openpar,
-                _PREC_LOGICAL_OR, _PREC_LOGICAL_AND)
-        from pymbolic.primitives import (
-                Comparison, LogicalAnd, LogicalOr)
+            _PREC_CALL, _PREC_COMPARISON, _PREC_LOGICAL_AND, _PREC_LOGICAL_OR,
+            _openpar)
+        from pymbolic.primitives import Comparison, LogicalAnd, LogicalOr
 
         next_tag = pstate.next_tag()
         if next_tag is _openpar and _PREC_CALL > min_precedence:
@@ -351,7 +348,7 @@ class ComplexCCodeMapper(CCodeMapperBase):
             complexes = [child for child in expr.children
                     if "c" == self.infer_type(child).kind]
 
-            from pymbolic.mapper.stringifier import PREC_SUM, PREC_NONE
+            from pymbolic.mapper.stringifier import PREC_NONE, PREC_SUM
             real_sum = self.join_rec(" + ", reals, PREC_SUM)
 
             if len(complexes) == 1:
@@ -387,7 +384,7 @@ class ComplexCCodeMapper(CCodeMapperBase):
             complexes = [child for child in expr.children
                     if "c" == self.infer_type(child).kind]
 
-            from pymbolic.mapper.stringifier import PREC_PRODUCT, PREC_NONE
+            from pymbolic.mapper.stringifier import PREC_NONE, PREC_PRODUCT
             real_prd = self.join_rec("*", reals, PREC_PRODUCT)
 
             if len(complexes) == 1:
@@ -489,7 +486,7 @@ class CCodeMapper(ComplexCCodeMapper):
         if isinstance(idx, tuple) and len(idx) == 1:
             idx, = idx
 
-        from pymbolic.mapper.stringifier import PREC_NONE, PREC_CALL
+        from pymbolic.mapper.stringifier import PREC_CALL, PREC_NONE
         return self.parenthesize_if_needed(
                 self.format("%s[%s%s]",
                     self.scope.translate_var_name(expr.aggregate.name),

{pyopencl-2023.1.4 → pyopencl-2024.2}/doc/Makefile

@@ -3,7 +3,7 @@
 
 # You can set these variables from the command line, and also
 # from the environment for the first two.
-SPHINXOPTS    ?=
+SPHINXOPTS    ?= -W -n
 SPHINXBUILD   ?= python $(shell which sphinx-build)
 SOURCEDIR     = .
 BUILDDIR      = _build

{pyopencl-2023.1.4 → pyopencl-2024.2}/doc/conf.py

@@ -1,5 +1,6 @@
 from urllib.request import urlopen
 
+
 _conf_url = \
         "https://raw.githubusercontent.com/inducer/sphinxconfig/main/sphinxconfig.py"
 with urlopen(_conf_url) as _inf:
@@ -21,4 +22,5 @@ intersphinx_mapping = {
     "python": ("https://docs.python.org/3", None),
     "numpy": ("https://numpy.org/doc/stable/", None),
     "mako": ("https://docs.makotemplates.org/en/latest", None),
+    "pytools": ("https://documen.tician.de/pytools", None),
 }

{pyopencl-2023.1.4 → pyopencl-2024.2}/doc/make_constants.py

@@ -22,6 +22,7 @@ THE SOFTWARE.
 
 import pyopencl as cl
 
+
 fission = ("cl_ext_device_fission", "2011.1")
 nv_devattr = ("cl_nv_device_attribute_query", "0.92")
 gl_sharing = ("cl_khr_gl_sharing", "0.92")
@@ -548,6 +549,7 @@ if not cl.have_gl():
 
 import inspect
 
+
 CONSTANT_CLASSES = [
         getattr(cl, name) for name in dir(cl)
         if inspect.isclass(getattr(cl, name))

{pyopencl-2023.1.4 → pyopencl-2024.2}/doc/runtime_platform.rst

@@ -104,6 +104,8 @@ Device
 
         .. versionadded:: 2020.3
 
+.. autofunction:: choose_devices
+
 Context
 -------
 
@@ -122,9 +124,9 @@ Context
 
     .. note::
 
-        Calling the constructor with no arguments will fail for recent
-        CL drivers that support the OpenCL ICD. If you want similar,
-        just-give-me-a-context-already behavior, we recommend
+        Calling the constructor with no arguments will fail for
+        CL drivers that support the OpenCL ICD (which applies to most modern systems).
+        If you want similar, just-give-me-a-context-already behavior, we recommend
         :func:`create_some_context`. See, e.g. this
         `explanation by AMD
         <https://web.archive.org/web/20101114195033/https://developer.amd.com/support/KnowledgeBase/Lists/KnowledgeBase/DispForm.aspx?ID=71>`__.
@@ -179,13 +181,4 @@ Context
 
     |comparable|
 
-.. function:: create_some_context(interactive=True, answers=None, cache_dir=None)
-
-    Create a :class:`Context` 'somehow'.
-
-    If multiple choices for platform and/or device exist, *interactive*
-    is True, and *sys.stdin.isatty()* is also True,
-    then the user is queried about which device should be chosen.
-    Otherwise, a device is chosen in an implementation-defined manner.
-
-
+.. autofunction:: create_some_context

{pyopencl-2023.1.4 → pyopencl-2024.2}/doc/runtime_program.rst

@@ -15,10 +15,10 @@ Program
     constructed with source. (It will depend on the ICD in use how much
     compilation work is saved by this.)
     By setting the environment variable :envvar:`PYOPENCL_NO_CACHE` to any
-    non-empty value, this caching is suppressed. No additional in-memory
-    caching is performed. To retain the compiled version of a kernel
-    in memory, simply retain the :class:`Program` and/or :class:`Kernel`
-    objects.
+    string that :func:`pytools.strtobool` evaluates as ``True``, this caching
+    is suppressed. No additional in-memory caching is performed. To retain the
+    compiled version of a kernel in memory, simply retain the :class:`Program`
+    and/or :class:`Kernel` objects.
 
     PyOpenCL will also cache "invokers", which are short snippets of Python
     that are generated to accelerate passing arguments to and enqueuing
@@ -26,6 +26,12 @@ Program
 
     .. versionadded:: 2013.1
 
+.. envvar:: PYOPENCL_COMPILER_OUTPUT
+
+    When setting the environment variable :envvar:`PYOPENCL_COMPILER_OUTPUT`
+    to any string that :func:`pytools.strtobool` evaluates as ``True``,
+    PyOpenCL will show compiler messages emitted during program build.
+
 .. envvar:: PYOPENCL_BUILD_OPTIONS
 
     Any options found in the environment variable
@@ -101,8 +107,8 @@ Program
         Instead, either use the (recommended, stateless) calling interface::
 
             sum_knl = prg.sum
-            sum_knl(queue, a_np.shape, None, a_g, b_g, res_g)    
-            
+            sum_knl(queue, a_np.shape, None, a_g, b_g, res_g)
+
         or the long, stateful way around, if you prefer::
 
             sum_knl.set_args(a_g, b_g, res_g)

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/black-hole-accretion.py

@@ -32,13 +32,15 @@
 # export CL_CONFIG_USE_VECTORIZER=True
 # export CL_CONFIG_CPU_VECTORIZER_MODE=16
 
-import pyopencl as cl
-import numpy
-import time
-import sys
 import getopt
+import sys
+import time
 from socket import gethostname
 
+import numpy
+
+import pyopencl as cl
+
 
 def DictionariesAPI():
     PhysicsList = {"Einstein": 0, "Newton": 1}

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo-struct-reduce.py

@@ -1,4 +1,5 @@
 import numpy as np
+
 import pyopencl as cl
 
 
@@ -55,10 +56,14 @@ preamble = mmc_c_decl + r"""//CL//
     """
 
 from pyopencl.clrandom import rand as clrand
+
+
 a_gpu = clrand(queue, (20000,), dtype=np.int32, a=0, b=10**6)
 a = a_gpu.get()
 
 from pyopencl.reduction import ReductionKernel
+
+
 red = ReductionKernel(ctx, mmc_dtype,
         neutral="mmc_neutral()",
         reduce_expr="agg_mmc(a, b)", map_expr="mmc_from_scalar(x[i])",

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo.py

@@ -1,10 +1,13 @@
 #!/usr/bin/env python
 
 import numpy as np
+
 import pyopencl as cl
 
-a_np = np.random.rand(50000).astype(np.float32)
-b_np = np.random.rand(50000).astype(np.float32)
+
+rng = np.random.default_rng()
+a_np = rng.random(50000, dtype=np.float32)
+b_np = rng.random(50000, dtype=np.float32)
 
 ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)
@@ -30,6 +33,7 @@ res_np = np.empty_like(a_np)
 cl.enqueue_copy(queue, res_np, res_g)
 
 # Check on CPU with Numpy:
-print(res_np - (a_np + b_np))
-print(np.linalg.norm(res_np - (a_np + b_np)))
+error_np = res_np - (a_np + b_np)
+print(f"Error:\n{error_np}")
+print(f"Norm: {np.linalg.norm(error_np):.16e}")
 assert np.allclose(res_np, a_np + b_np)

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo_array.py

@@ -1,10 +1,13 @@
+import numpy as np
+import numpy.linalg as la
+
 import pyopencl as cl
 import pyopencl.array as cl_array
-import numpy
-import numpy.linalg as la
 
-a = numpy.random.rand(50000).astype(numpy.float32)
-b = numpy.random.rand(50000).astype(numpy.float32)
+
+rng = np.random.default_rng()
+a = rng.random(50000, dtype=np.float32)
+b = rng.random(50000, dtype=np.float32)
 
 ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)
@@ -26,3 +29,4 @@ knl = prg.sum  # Use this Kernel object for repeated calls
 knl(queue, a.shape, None, a_dev.data, b_dev.data, dest_dev.data)
 
 print(la.norm((dest_dev - (a_dev+b_dev)).get()))
+assert np.allclose(dest_dev.get(), (a_dev + b_dev).get())

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo_array_svm.py

@@ -1,12 +1,15 @@
+import numpy as np
+
 import pyopencl as cl
 import pyopencl.array as cl_array
 from pyopencl.tools import SVMAllocator, SVMPool
-import numpy as np
+
 
 n = 50000
-a = np.random.rand(n).astype(np.float32)
-b = np.random.rand(n).astype(np.float32)
 
+rng = np.random.default_rng()
+a = rng.random(n, dtype=np.float32)
+b = rng.random(n, dtype=np.float32)
 
 ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)
@@ -30,4 +33,5 @@ prg = cl.Program(ctx, """
 knl = prg.sum
 knl(queue, a.shape, None, a_dev.data, b_dev.data, dest_dev.data)
 
-np.testing.assert_allclose(dest_dev.get(), (a_dev+b_dev).get())
+print(np.linalg.norm((dest_dev - (a_dev + b_dev)).get()))
+assert np.allclose(dest_dev.get(), (a_dev + b_dev).get())

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo_elementwise.py

@@ -1,13 +1,15 @@
-#!/usr/bin/env python
-
 import numpy as np
+
 import pyopencl as cl
 import pyopencl.array
 from pyopencl.elementwise import ElementwiseKernel
 
+
 n = 10
-a_np = np.random.randn(n).astype(np.float32)
-b_np = np.random.randn(n).astype(np.float32)
+
+rng = np.random.default_rng()
+a_np = rng.random(n, dtype=np.float32)
+b_np = rng.random(n, dtype=np.float32)
 
 ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)

{pyopencl-2023.1.4 → pyopencl-2024.2}/examples/demo_elementwise_complex.py

@@ -1,18 +1,24 @@
+import numpy as np
+import numpy.linalg as la
+
 import pyopencl as cl
 import pyopencl.array as cl_array
-import numpy
-import numpy.linalg as la
+from pyopencl.elementwise import ElementwiseKernel
+
 
 ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)
 
 n = 10
+
+rng = np.random.default_rng()
 a_gpu = cl_array.to_device(queue,
-        (numpy.random.randn(n) + 1j*numpy.random.randn(n)).astype(numpy.complex64))
+        rng.standard_normal(n, dtype=np.float32)
+        + 1j*rng.standard_normal(n, dtype=np.float32))
 b_gpu = cl_array.to_device(queue,
-        (numpy.random.randn(n) + 1j*numpy.random.randn(n)).astype(numpy.complex64))
+        rng.standard_normal(n, dtype=np.float32)
+        + 1j*rng.standard_normal(n, dtype=np.float32))
 
-from pyopencl.elementwise import ElementwiseKernel
 complex_prod = ElementwiseKernel(ctx,
         "float a, "
         "cfloat_t *x, "
@@ -39,7 +45,7 @@ real_part = ElementwiseKernel(ctx,
 c_gpu = cl_array.empty_like(a_gpu)
 complex_prod(5, a_gpu, b_gpu, c_gpu)
 
-c_gpu_real = cl_array.empty(queue, len(a_gpu), dtype=numpy.float32)
+c_gpu_real = cl_array.empty(queue, len(a_gpu), dtype=np.float32)
 real_part(c_gpu, c_gpu_real)
 print(c_gpu.get().real - c_gpu_real.get())