Skip to content
Snippets Groups Projects

Compare revisions

Changes are shown as if the source revision was being merged into the target revision. Learn more about comparing revisions.

Source

Select target project
No results found

Target

Select target project
  • inducer/arraycontext
  • kaushikcfd/arraycontext
  • fikl2/arraycontext
3 results
Show changes
Commits on Source (330)
230 additional commits have been omitted to prevent performance issues.
Hide whitespace changes
Inline Side-by-side
Showing
with 3013 additions and 743 deletions
.github/dependabot.yml 0 → 100644
View file @ 21a89f1b
version: 2
updates:
# Set update schedule for GitHub Actions
- package-ecosystem: "github-actions"
directory: "/"
schedule:
interval: "weekly"
# vim: sw=4
.github/workflows/autopush.yml
View file @ 21a89f1b
......@@ -7,14 +7,15 @@ on:
jobs:
autopush:
name: Automatic push to gitlab.tiker.net
if: startsWith(github.repository, 'inducer/')
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- run: |
mkdir ~/.ssh && echo -e "Host gitlab.tiker.net\n\tStrictHostKeyChecking no\n" >> ~/.ssh/config
eval $(ssh-agent) && echo "$GITLAB_AUTOPUSH_KEY" | ssh-add -
git fetch --unshallow
git push "git@gitlab.tiker.net:inducer/$(basename $GITHUB_REPOSITORY).git" main
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
mirror_github_to_gitlab
env:
GITLAB_AUTOPUSH_KEY: ${{ secrets.GITLAB_AUTOPUSH_KEY }}
......
.github/workflows/ci.yml
View file @ 21a89f1b
......@@ -7,27 +7,35 @@ on:
schedule:
- cron: '17 3 * * 0'
concurrency:
group: ${{ github.head_ref || github.ref_name }}
cancel-in-progress: true
jobs:
flake8:
name: Flake8
typos:
name: Typos
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: crate-ci/typos@master
ruff:
name: Ruff
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
-
uses: actions/setup-python@v1
with:
# matches compat target in setup.py
python-version: '3.6'
uses: actions/setup-python@v5
- name: "Main Script"
run: |
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/prepare-and-run-flake8.sh
. ./prepare-and-run-flake8.sh "$(basename $GITHUB_REPOSITORY)" test examples
pip install ruff
ruff check
pylint:
name: Pylint
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: "Main Script"
run: |
USE_CONDA_BUILD=1
......@@ -38,28 +46,25 @@ jobs:
name: Mypy
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
-
uses: actions/setup-python@v1
with:
python-version: '3.x'
- uses: actions/checkout@v4
- name: "Main Script"
run: |
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
build_py_project_in_conda_env
python -m pip install mypy
python -m pip install mypy pytest
./run-mypy.sh
pytest3_pocl:
name: Pytest Conda Py3 POCL
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: "Main Script"
run: |
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/ci-support.sh
. ./ci-support.sh
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
build_py_project_in_conda_env
test_py_project
......@@ -67,7 +72,7 @@ jobs:
name: Pytest Conda Py3 Intel
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: "Main Script"
run: |
curl -L -O https://raw.githubusercontent.com/illinois-scicomp/machine-shop-maintenance/main/install-intel-icd.sh
......@@ -88,7 +93,7 @@ jobs:
name: Examples Conda Py3
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: "Main Script"
run: |
export MPLBACKEND=Agg
......@@ -100,15 +105,15 @@ jobs:
name: Documentation
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
-
uses: actions/setup-python@v1
uses: actions/setup-python@v5
with:
python-version: '3.x'
- name: "Main Script"
run: |
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/ci-support.sh
. ci-support.sh
curl -L -O https://tiker.net/ci-support-v0
. ci-support-v0
build_py_project_in_conda_env
conda install graphviz
......@@ -118,46 +123,21 @@ jobs:
downstream_tests:
strategy:
matrix:
#downstream_project: [meshmode, grudge, pytential, mirgecom]
downstream_project: [meshmode, grudge, mirgecom]
downstream_project: [meshmode, grudge, mirgecom, mirgecom_examples]
fail-fast: false
name: Tests for downstream project ${{ matrix.downstream_project }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: "Main Script"
env:
DOWNSTREAM_PROJECT: ${{ matrix.downstream_project }}
run: |
if test "$DOWNSTREAM_PROJECT" = "mirgecom"; then
git clone "https://github.com/illinois-ceesd/$DOWNSTREAM_PROJECT.git"
else
git clone "https://github.com/inducer/$DOWNSTREAM_PROJECT.git"
fi
cd "$DOWNSTREAM_PROJECT"
echo "*** $DOWNSTREAM_PROJECT version: $(git rev-parse --short HEAD)"
# Use this version of arraycontext instead of what downstream would install
sed -i "/egg=arraycontext/ c git+file://$(readlink -f ..)#egg=arraycontext" requirements.txt
# Avoid slow or complicated tests in downstream projects
export PYTEST_ADDOPTS="-k 'not (slowtest or octave or mpi)'"
if test "$DOWNSTREAM_PROJECT" = "mirgecom"; then
# can't turn off MPI in mirgecom
sudo apt-get update
sudo apt-get install openmpi-bin libopenmpi-dev
export CONDA_ENVIRONMENT=conda-env.yml
export CISUPPORT_PARALLEL_PYTEST=no
else
sed -i "/mpi4py/ d" requirements.txt
fi
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
build_py_project_in_conda_env
test_py_project
test_downstream "$DOWNSTREAM_PROJECT"
if [[ "$DOWNSTREAM_PROJECT" = "meshmode" ]]; then
python ../examples/simple-dg.py --lazy
fi
......
.gitlab-ci.yml
View file @ 21a89f1b
Python 3 POCL:
script: |
export PY_EXE=python3
export PYOPENCL_TEST=portable:pthread
# cython is here because pytential (for now, for TS) depends on it
export EXTRA_INSTALL="pybind11 cython numpy mako mpi4py oct2py"
export PYOPENCL_TEST=portable:cpu
export EXTRA_INSTALL="jax[cpu]"
export JAX_PLATFORMS=cpu
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/build-and-test-py-project.sh
. ./build-and-test-py-project.sh
tags:
......@@ -18,12 +17,30 @@ Python 3 POCL:
Python 3 Nvidia Titan V:
script: |
export PY_EXE=python3
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
export PYOPENCL_TEST=nvi:titan
export EXTRA_INSTALL="pybind11 cython numpy mako oct2py"
# cython is here because pytential (for now, for TS) depends on it
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/build-and-test-py-project.sh
. ./build-and-test-py-project.sh
build_py_project_in_venv
pip install --upgrade "jax[cuda]" -f https://storage.googleapis.com/jax-releases/jax_releases.html
test_py_project
tags:
- python3
- nvidia-titan-v
except:
- tags
artifacts:
reports:
junit: test/pytest.xml
Python 3 POCL Nvidia Titan V:
script: |
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
export PYOPENCL_TEST=port:titan
build_py_project_in_venv
test_py_project
tags:
- python3
- nvidia-titan-v
......@@ -36,9 +53,7 @@ Python 3 Nvidia Titan V:
Python 3 POCL Examples:
script:
- test -n "$SKIP_EXAMPLES" && exit
- export PY_EXE=python3
- export PYOPENCL_TEST=portable:pthread
- export EXTRA_INSTALL="pybind11 numpy mako"
- export PYOPENCL_TEST=portable:cpu
- curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/build-py-project-and-run-examples.sh
- ". ./build-py-project-and-run-examples.sh"
tags:
......@@ -50,6 +65,11 @@ Python 3 POCL Examples:
Python 3 Conda:
script: |
export PYOPENCL_TEST=portable:cpu
# Avoid crashes like https://gitlab.tiker.net/inducer/arraycontext/-/jobs/536021
sed -i 's/jax/jax !=0.4.6/' .test-conda-env-py3.yml
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/build-and-test-py-project-within-miniconda.sh
. ./build-and-test-py-project-within-miniconda.sh
tags:
......@@ -61,26 +81,24 @@ Python 3 Conda:
Documentation:
script: |
EXTRA_INSTALL="pybind11 cython numpy"
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/build-docs.sh
CI_SUPPORT_SPHINX_VERSION_SPECIFIER=">=4.0"
. ./build-docs.sh
tags:
- python3
Flake8:
Ruff:
script:
- curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/main/prepare-and-run-flake8.sh
- . ./prepare-and-run-flake8.sh "$CI_PROJECT_NAME" test examples
- pipx install ruff
- ruff check
tags:
- python3
- docker-runner
except:
- tags
Pylint:
script: |
export PY_EXE=python3
EXTRA_INSTALL="pybind11 numpy mako"
EXTRA_INSTALL="jax[cpu]"
curl -L -O https://gitlab.tiker.net/inducer/ci-support/raw/master/prepare-and-run-pylint.sh
. ./prepare-and-run-pylint.sh "$CI_PROJECT_NAME" examples/*.py test/test_*.py
tags:
......@@ -90,12 +108,30 @@ Pylint:
Mypy:
script: |
EXTRA_INSTALL="mypy pytest"
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
build_py_project_in_venv
python -m pip install mypy
./run-mypy.sh
tags:
- python3
except:
- tags
Downstream:
parallel:
matrix:
- DOWNSTREAM_PROJECT: [meshmode, grudge, mirgecom, mirgecom_examples]
tags:
- large-node
- "docker-runner"
script: |
curl -L -O https://tiker.net/ci-support-v0
. ./ci-support-v0
test_downstream "$DOWNSTREAM_PROJECT"
if [[ "$DOWNSTREAM_PROJECT" = "meshmode" ]]; then
python ../examples/simple-dg.py --lazy
fi
.test-conda-env-py3.yml
View file @ 21a89f1b
......@@ -8,8 +8,10 @@ dependencies:
- git
- libhwloc=2
- numpy
- pocl
# pocl 3.1 required for full SVM functionality
- pocl>=3.1
- mako
- pyopencl
- islpy
- pip
- jax
MANIFEST.in deleted 100644 → 0
View file @ f74fb9e9
include doc/*.rst
include doc/conf.py
include doc/make.bat
include doc/Makefile
include examples/*.py
README.rst
View file @ 21a89f1b
......@@ -7,7 +7,7 @@ arraycontext: Choose your favorite ``numpy``-workalike
.. image:: https://github.com/inducer/arraycontext/workflows/CI/badge.svg
:alt: Github Build Status
:target: https://github.com/inducer/arraycontext/actions?query=branch%3Amain+workflow%3ACI
.. image:: https://badge.fury.io/py/arraycontext.png
.. image:: https://badge.fury.io/py/arraycontext.svg
:alt: Python Package Index Release Page
:target: https://pypi.org/project/arraycontext/
......@@ -17,7 +17,9 @@ implementations for:
- numpy
- `PyOpenCL <https://documen.tician.de/pyopencl/array.html>`__
- `JAX <https://jax.readthedocs.io/en/latest/>`__
- `Pytato <https://documen.tician.de/pytato>`__ (for lazy/deferred evaluation)
with backends for ``pyopencl`` and ``jax``.
- Debugging
- Profiling
......
arraycontext/__init__.py
View file @ 21a89f1b
......@@ -2,6 +2,7 @@
An array context is an abstraction that helps you dispatch between multiple
implementations of :mod:`numpy`-like :math:`n`-dimensional arrays.
"""
from __future__ import annotations
__copyright__ = """
......@@ -28,75 +29,140 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
import sys
from .context import ArrayContext
from .transform_metadata import (CommonSubexpressionTag,
ElementwiseMapKernelTag)
# deprecated, remove in 2022.
from .metadata import _FirstAxisIsElementsTag
from .container import (
ArrayContainer,
is_array_container, is_array_container_type,
get_container_context, get_container_context_recursively,
serialize_container, deserialize_container)
from .container.arithmetic import with_container_arithmetic
ArithArrayContainer,
ArrayContainer,
ArrayContainerT,
NotAnArrayContainerError,
SerializationKey,
SerializedContainer,
deserialize_container,
get_container_context_opt,
get_container_context_recursively,
get_container_context_recursively_opt,
is_array_container,
is_array_container_type,
register_multivector_as_array_container,
serialize_container,
)
from .container.arithmetic import (
with_container_arithmetic,
)
from .container.dataclass import dataclass_array_container
from .container.traversal import (
map_array_container,
multimap_array_container,
rec_map_array_container,
rec_multimap_array_container,
mapped_over_array_containers,
multimapped_over_array_containers,
rec_map_reduce_array_container,
rec_multimap_reduce_array_container,
thaw, freeze,
from_numpy, to_numpy)
flat_size_and_dtype,
flatten,
freeze,
from_numpy,
map_array_container,
map_reduce_array_container,
mapped_over_array_containers,
multimap_array_container,
multimap_reduce_array_container,
multimapped_over_array_containers,
outer,
rec_map_array_container,
rec_map_reduce_array_container,
rec_multimap_array_container,
rec_multimap_reduce_array_container,
stringify_array_container_tree,
thaw,
to_numpy,
unflatten,
with_array_context,
)
from .context import (
Array,
ArrayContext,
ArrayOrArithContainer,
ArrayOrArithContainerOrScalar,
ArrayOrArithContainerOrScalarT,
ArrayOrArithContainerT,
ArrayOrContainer,
ArrayOrContainerOrScalar,
ArrayOrContainerOrScalarT,
ArrayOrContainerT,
ArrayT,
Scalar,
ScalarLike,
tag_axes,
)
from .impl.jax import EagerJAXArrayContext
from .impl.numpy import NumpyArrayContext
from .impl.pyopencl import PyOpenCLArrayContext
from .impl.pytato import PytatoPyOpenCLArrayContext
from .pytest import (
PytestPyOpenCLArrayContextFactory,
pytest_generate_tests_for_array_contexts,
pytest_generate_tests_for_pyopencl_array_context)
from .impl.pytato import PytatoJAXArrayContext, PytatoPyOpenCLArrayContext
from .loopy import make_loopy_program
from .pytest import (
PytestArrayContextFactory,
PytestPyOpenCLArrayContextFactory,
pytest_generate_tests_for_array_contexts,
)
from .transform_metadata import CommonSubexpressionTag, ElementwiseMapKernelTag
__all__ = (
"ArrayContext",
"CommonSubexpressionTag",
"ElementwiseMapKernelTag",
"ArrayContainer",
"is_array_container", "is_array_container_type",
"get_container_context", "get_container_context_recursively",
"serialize_container", "deserialize_container",
"with_container_arithmetic",
"dataclass_array_container",
"map_array_container", "multimap_array_container",
"rec_map_array_container", "rec_multimap_array_container",
"mapped_over_array_containers",
"multimapped_over_array_containers",
"rec_map_reduce_array_container", "rec_multimap_reduce_array_container",
"thaw", "freeze",
"from_numpy", "to_numpy",
"PyOpenCLArrayContext", "PytatoPyOpenCLArrayContext",
"make_loopy_program",
"PytestPyOpenCLArrayContextFactory",
"pytest_generate_tests_for_array_contexts",
"pytest_generate_tests_for_pyopencl_array_context"
)
"ArithArrayContainer",
"Array",
"ArrayContainer",
"ArrayContainerT",
"ArrayContext",
"ArrayOrArithContainer",
"ArrayOrArithContainerOrScalar",
"ArrayOrArithContainerOrScalarT",
"ArrayOrArithContainerT",
"ArrayOrContainer",
"ArrayOrContainerOrScalar",
"ArrayOrContainerOrScalarT",
"ArrayOrContainerT",
"ArrayT",
"CommonSubexpressionTag",
"EagerJAXArrayContext",
"ElementwiseMapKernelTag",
"NotAnArrayContainerError",
"NumpyArrayContext",
"PyOpenCLArrayContext",
"PytatoJAXArrayContext",
"PytatoPyOpenCLArrayContext",
"PytestArrayContextFactory",
"PytestPyOpenCLArrayContextFactory",
"Scalar",
"ScalarLike",
"SerializationKey",
"SerializedContainer",
"dataclass_array_container",
"deserialize_container",
"flat_size_and_dtype",
"flatten",
"freeze",
"from_numpy",
"get_container_context_opt",
"get_container_context_recursively",
"get_container_context_recursively_opt",
"is_array_container",
"is_array_container_type",
"make_loopy_program",
"map_array_container",
"map_reduce_array_container",
"mapped_over_array_containers",
"multimap_array_container",
"multimap_reduce_array_container",
"multimapped_over_array_containers",
"outer",
"pytest_generate_tests_for_array_contexts",
"rec_map_array_container",
"rec_map_reduce_array_container",
"rec_multimap_array_container",
"rec_multimap_reduce_array_container",
"register_multivector_as_array_container",
"serialize_container",
"stringify_array_container_tree",
"tag_axes",
"thaw",
"to_numpy",
"unflatten",
"with_array_context",
"with_container_arithmetic",
)
# {{{ deprecation handling
......@@ -113,29 +179,24 @@ def _deprecated_acf():
_depr_name_to_replacement_and_obj = {
"FirstAxisIsElementsTag":
("meshmode.transform_metadata.FirstAxisIsElementsTag",
_FirstAxisIsElementsTag),
"_acf":
("<no replacement yet>", _deprecated_acf),
"get_container_context": (
"get_container_context_opt",
get_container_context_opt, 2022),
}
if sys.version_info >= (3, 7):
def __getattr__(name):
replacement_and_obj = _depr_name_to_replacement_and_obj.get(name, None)
if replacement_and_obj is not None:
replacement, obj = replacement_and_obj
from warnings import warn
warn(f"'arraycontext.{name}' is deprecated. "
f"Use '{replacement}' instead. "
f"'arraycontext.{name}' will continue to work until 2022.",
DeprecationWarning, stacklevel=2)
return obj
else:
raise AttributeError(name)
else:
FirstAxisIsElementsTag = _FirstAxisIsElementsTag
_acf = _deprecated_acf
def __getattr__(name):
replacement_and_obj = _depr_name_to_replacement_and_obj.get(name)
if replacement_and_obj is not None:
replacement, obj, year = replacement_and_obj
from warnings import warn
warn(f"'arraycontext.{name}' is deprecated. "
f"Use '{replacement}' instead. "
f"'arraycontext.{name}' will continue to work until {year}.",
DeprecationWarning, stacklevel=2)
return obj
else:
raise AttributeError(name)
# }}}
......
arraycontext/container/__init__.py
View file @ 21a89f1b
......@@ -3,25 +3,58 @@
"""
.. currentmodule:: arraycontext
.. autoclass:: ArrayContainer
.. autoclass:: ArithArrayContainer
.. class:: ArrayContainerT
:canonical: arraycontext.container.ArrayContainerT
:class:`~typing.TypeVar` for array container-like objects.
A type variable with a lower bound of :class:`ArrayContainer`.
.. autoclass:: ArrayContainer
.. autoexception:: NotAnArrayContainerError
Serialization/deserialization
-----------------------------
.. autofunction:: is_array_container
.. autoclass:: SerializationKey
.. autoclass:: SerializedContainer
.. autofunction:: is_array_container_type
.. autofunction:: serialize_container
.. autofunction:: deserialize_container
Context retrieval
-----------------
.. autofunction:: get_container_context
.. autofunction:: get_container_context_opt
.. autofunction:: get_container_context_recursively
.. autofunction:: get_container_context_recursively_opt
:class:`~pymbolic.geometric_algebra.MultiVector` support
---------------------------------------------------------
.. autofunction:: register_multivector_as_array_container
.. currentmodule:: arraycontext.container
Canonical locations for type annotations
----------------------------------------
.. class:: ArrayContainerT
:canonical: arraycontext.ArrayContainerT
.. class:: ArrayOrContainerT
:canonical: arraycontext.ArrayOrContainerT
.. class:: SerializationKey
:canonical: arraycontext.SerializationKey
.. class:: SerializedContainer
:canonical: arraycontext.SerializedContainer
"""
from __future__ import annotations
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
......@@ -47,19 +80,30 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from collections.abc import Hashable, Sequence
from functools import singledispatch
from arraycontext.context import ArrayContext
from typing import Any, Iterable, Tuple, TypeVar, Optional
from typing import TYPE_CHECKING, Protocol, TypeAlias, TypeVar
# For use in singledispatch type annotations, because sphinx can't figure out
# what 'np' is.
import numpy
import numpy as np
from typing_extensions import Self
from arraycontext.context import ArrayContext, ArrayOrScalar
ArrayContainerT = TypeVar("ArrayContainerT")
if TYPE_CHECKING:
from pymbolic.geometric_algebra import MultiVector
from arraycontext import ArrayOrContainer
# {{{ ArrayContainer
class ArrayContainer:
r"""
A generic container for the array type supported by the
class ArrayContainer(Protocol):
"""
A protocol for generic containers of the array type supported by the
:class:`ArrayContext`.
The functionality required for the container to operated is supplied via
......@@ -70,13 +114,11 @@ class ArrayContainer:
of the array.
* :func:`deserialize_container` for deserialization, which constructs a
container from a set of components.
* :func:`get_container_context` retrieves the :class:`ArrayContext` from
* :func:`get_container_context_opt` retrieves the :class:`ArrayContext` from
a container, if it has one.
This allows enumeration of the component arrays in a container and the
construction of modified containers from an iterable of those component arrays.
:func:`is_array_container` will return *True* for types that have
a container serialization function registered.
Packages may register their own types as array containers. They must not
register other types (e.g. :class:`list`) as array containers.
......@@ -92,67 +134,138 @@ class ArrayContainer:
.. note::
This class is used in type annotation. Inheriting from it confers no
special meaning or behavior.
This class is used in type annotation and as a marker of array container
attributes for :func:`~arraycontext.dataclass_array_container`.
As a protocol, it is not intended as a superclass.
"""
# Array containers do not need to have any particular features, so this
# protocol is deliberately empty.
# This *is* used as a type annotation in dataclasses that are processed
# by dataclass_array_container, where it's used to recognize attributes
# that are container-typed.
class ArithArrayContainer(ArrayContainer, Protocol):
"""
A sub-protocol of :class:`ArrayContainer` that supports basic arithmetic.
"""
# This is loose and permissive, assuming that any array can be added
# to any container. The alternative would be to plaster type-ignores
# on all those uses. Achieving typing precision on what broadcasting is
# allowable seems like a huge endeavor and is likely not feasible without
# a mypy plugin. Maybe some day? -AK, November 2024
def __neg__(self) -> Self: ...
def __abs__(self) -> Self: ...
def __add__(self, other: ArrayOrScalar | Self) -> Self: ...
def __radd__(self, other: ArrayOrScalar | Self) -> Self: ...
def __sub__(self, other: ArrayOrScalar | Self) -> Self: ...
def __rsub__(self, other: ArrayOrScalar | Self) -> Self: ...
def __mul__(self, other: ArrayOrScalar | Self) -> Self: ...
def __rmul__(self, other: ArrayOrScalar | Self) -> Self: ...
def __truediv__(self, other: ArrayOrScalar | Self) -> Self: ...
def __rtruediv__(self, other: ArrayOrScalar | Self) -> Self: ...
ArrayContainerT = TypeVar("ArrayContainerT", bound=ArrayContainer)
class NotAnArrayContainerError(TypeError):
""":class:`TypeError` subclass raised when an array container is expected."""
SerializationKey: TypeAlias = Hashable
SerializedContainer: TypeAlias = Sequence[tuple[SerializationKey, "ArrayOrContainer"]]
@singledispatch
def serialize_container(ary: ArrayContainer) -> Iterable[Tuple[Any, Any]]:
r"""Serialize the array container into an iterable over its components.
def serialize_container(
ary: ArrayContainer) -> SerializedContainer:
r"""Serialize the array container into a sequence over its components.
The order of the components and their identifiers are entirely under
the control of the container class.
the control of the container class. However, the order is required to be
deterministic, i.e. two calls to :func:`serialize_container` on
array containers of the same types with the same number of
sub-arrays must result in a sequence with the keys in the same
order.
If *ary* is mutable, the serialization function is not required to ensure
that the serialization result reflects the array state at the time of the
call to :func:`serialize_container`.
:returns: an :class:`Iterable` of 2-tuples where the first
:returns: a :class:`Sequence` of 2-tuples where the first
entry is an identifier for the component and the second entry
is an array-like component of the :class:`ArrayContainer`.
Components can themselves be :class:`ArrayContainer`\ s, allowing
for arbitrarily nested structures. The identifiers need to be hashable
but are otherwise treated as opaque.
"""
raise NotImplementedError(type(ary).__name__)
raise NotAnArrayContainerError(
f"'{type(ary).__name__}' cannot be serialized as a container")
@singledispatch
def deserialize_container(template: Any, iterable: Iterable[Tuple[Any, Any]]) -> Any:
"""Deserialize an iterable into an array container.
def deserialize_container(
template: ArrayContainerT,
serialized: SerializedContainer) -> ArrayContainerT:
"""Deserialize a sequence into an array container following a *template*.
:param template: an instance of an existing object that
can be used to aid in the deserialization. For a similar choice
see :attr:`~numpy.class.__array_finalize__`.
:param iterable: an iterable that mirrors the output of
:param serialized: a sequence that mirrors the output of
:meth:`serialize_container`.
"""
raise NotImplementedError(type(template).__name__)
raise NotAnArrayContainerError(
f"'{type(template).__name__}' cannot be deserialized as a container")
def is_array_container_type(cls: type) -> bool:
"""
:returns: *True* if the type *cls* has a registered implementation of
:func:`serialize_container`, or if it is an :class:`ArrayContainer`.
.. warning::
Not all instances of a type that this function labels an array container
must automatically be array containers. For example, while this
function will say that :class:`numpy.ndarray` is an array container
type, only object arrays *actually are* array containers.
"""
assert isinstance(cls, type), f"must pass a {type!r}, not a '{cls!r}'"
return (
cls is ArrayContainer
or (serialize_container.dispatch(cls)
is not serialize_container.__wrapped__)) # type:ignore[attr-defined]
def is_array_container(ary: Any) -> bool:
def is_array_container(ary: object) -> bool:
"""
:returns: *True* if the instance *ary* has a registered implementation of
:func:`serialize_container`.
"""
from warnings import warn
warn("is_array_container is deprecated and will be removed in 2022. "
"If you must know precisely whether something is an array container, "
"try serializing it and catch NotAnArrayContainerError. For a "
"cheaper option, see is_array_container_type.",
DeprecationWarning, stacklevel=2)
return (serialize_container.dispatch(ary.__class__)
is not serialize_container.__wrapped__) # type:ignore[attr-defined]
is not serialize_container.__wrapped__ # type:ignore[attr-defined]
# numpy values with scalar elements aren't array containers
and not (isinstance(ary, np.ndarray)
and ary.dtype.kind != "O")
)
@singledispatch
def get_container_context(ary: ArrayContainer) -> Optional[ArrayContext]:
def get_container_context_opt(ary: ArrayContainer) -> ArrayContext | None:
"""Retrieves the :class:`ArrayContext` from the container, if any.
This function is not recursive, so it will only search at the root level
......@@ -167,10 +280,11 @@ def get_container_context(ary: ArrayContainer) -> Optional[ArrayContext]:
# {{{ object arrays as array containers
@serialize_container.register(np.ndarray)
def _serialize_ndarray_container(ary: np.ndarray) -> Iterable[Tuple[Any, Any]]:
def _serialize_ndarray_container(
ary: numpy.ndarray) -> SerializedContainer:
if ary.dtype.char != "O":
raise ValueError(
f"only object arrays are supported, given dtype '{ary.dtype}'")
raise NotAnArrayContainerError(
f"cannot serialize '{type(ary).__name__}' with dtype '{ary.dtype}'")
# special-cased for speed
if ary.ndim == 1:
......@@ -181,20 +295,22 @@ def _serialize_ndarray_container(ary: np.ndarray) -> Iterable[Tuple[Any, Any]]:
for j in range(ary.shape[1])
]
else:
return np.ndenumerate(ary)
return list(np.ndenumerate(ary))
@deserialize_container.register(np.ndarray)
def _deserialize_ndarray_container(
template: np.ndarray,
iterable: Iterable[Tuple[Any, Any]]) -> np.ndarray:
# https://github.com/python/mypy/issues/13040
def _deserialize_ndarray_container( # type: ignore[misc]
template: numpy.ndarray,
serialized: SerializedContainer) -> numpy.ndarray:
# disallow subclasses
assert type(template) is np.ndarray
assert template.dtype.char == "O"
result = type(template)(template.shape, dtype=object)
for i, subary in iterable:
result[i] = subary
for i, subary in serialized:
# FIXME: numpy annotations don't seem to handle object arrays very well
result[i] = subary # type: ignore[call-overload]
return result
......@@ -203,37 +319,101 @@ def _deserialize_ndarray_container(
# {{{ get_container_context_recursively
def get_container_context_recursively(ary: Any) -> Optional[ArrayContext]:
def get_container_context_recursively_opt(
ary: ArrayContainer) -> ArrayContext | None:
"""Walks the :class:`ArrayContainer` hierarchy to find an
:class:`ArrayContext` associated with it.
If different components that have different array contexts are found at
any level, an assertion error is raised.
"""
actx = None
if not is_array_container(ary):
return actx
Returns *None* if no array context was found.
"""
# try getting the array context directly
actx = get_container_context(ary)
actx = get_container_context_opt(ary)
if actx is not None:
return actx
for _, subary in serialize_container(ary):
context = get_container_context_recursively(subary)
if context is None:
continue
try:
iterable = serialize_container(ary)
except NotAnArrayContainerError:
return actx
else:
for _, subary in iterable:
context = get_container_context_recursively_opt(subary)
if context is None:
continue
if not __debug__:
return context
elif actx is None:
actx = context
else:
assert actx is context
return actx
def get_container_context_recursively(ary: ArrayContainer) -> ArrayContext | None:
"""Walks the :class:`ArrayContainer` hierarchy to find an
:class:`ArrayContext` associated with it.
If different components that have different array contexts are found at
any level, an assertion error is raised.
if not __debug__:
return context
elif actx is None:
actx = context
else:
assert actx is context
Raises an error if no array container is found.
"""
actx = get_container_context_recursively_opt(ary)
if actx is None:
# raise ValueError("no array context was found")
from warnings import warn
warn("No array context was found. This will be an error starting in "
"July of 2022. If you would like the function to return "
"None if no array context was found, use "
"get_container_context_recursively_opt.",
DeprecationWarning, stacklevel=2)
return actx
# }}}
# {{{ MultiVector support, see pymbolic.geometric_algebra
# FYI: This doesn't, and never should, make arraycontext directly depend on pymbolic.
# (Though clearly there exists a dependency via loopy.)
def _serialize_multivec_as_container(mv: MultiVector) -> SerializedContainer:
return list(mv.data.items())
# FIXME: Ignored due to https://github.com/python/mypy/issues/13040
def _deserialize_multivec_as_container( # type: ignore[misc]
template: MultiVector,
serialized: SerializedContainer) -> MultiVector:
from pymbolic.geometric_algebra import MultiVector
return MultiVector(dict(serialized), space=template.space)
def _get_container_context_opt_from_multivec(mv: MultiVector) -> None:
return None
def register_multivector_as_array_container() -> None:
"""Registers :class:`~pymbolic.geometric_algebra.MultiVector` as an
:class:`ArrayContainer`. This function may be called multiple times. The
second and subsequent calls have no effect.
"""
from pymbolic.geometric_algebra import MultiVector
if MultiVector not in serialize_container.registry:
serialize_container.register(MultiVector)(_serialize_multivec_as_container)
deserialize_container.register(MultiVector)(
_deserialize_multivec_as_container)
get_container_context_opt.register(MultiVector)(
_get_container_context_opt_from_multivec)
assert MultiVector in serialize_container.registry
# }}}
# vim: foldmethod=marker
arraycontext/container/arithmetic.py
View file @ 21a89f1b
# mypy: disallow-untyped-defs
from __future__ import annotations
"""
__doc__ = """
.. currentmodule:: arraycontext
.. autofunction:: with_container_arithmetic
"""
import enum
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
......@@ -31,7 +33,10 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from typing import Any, Callable, Optional, Tuple, TypeVar, Union
import enum
from collections.abc import Callable
from typing import Any, TypeVar
from warnings import warn
import numpy as np
......@@ -86,7 +91,7 @@ _BINARY_OP_AND_DUNDER = [
]
def _format_unary_op_str(op_str: str, arg1: Union[Tuple[str, ...], str]) -> str:
def _format_unary_op_str(op_str: str, arg1: tuple[str, ...] | str) -> str:
if isinstance(arg1, tuple):
arg1_entry, arg1_container = arg1
return (f"{op_str.format(arg1_entry)} "
......@@ -96,20 +101,14 @@ def _format_unary_op_str(op_str: str, arg1: Union[Tuple[str, ...], str]) -> str:
def _format_binary_op_str(op_str: str,
arg1: Union[Tuple[str, ...], str],
arg2: Union[Tuple[str, ...], str]) -> str:
arg1: tuple[str, str] | str,
arg2: tuple[str, str] | str) -> str:
if isinstance(arg1, tuple) and isinstance(arg2, tuple):
import sys
if sys.version_info >= (3, 10):
strict_arg = ", strict=__debug__"
else:
strict_arg = ""
arg1_entry, arg1_container = arg1
arg2_entry, arg2_container = arg2
return (f"{op_str.format(arg1_entry, arg2_entry)} "
f"for {arg1_entry}, {arg2_entry} "
f"in zip({arg1_container}, {arg2_container}{strict_arg})")
f"in zip({arg1_container}, {arg2_container}, strict=__debug__)")
elif isinstance(arg1, tuple):
arg1_entry, arg1_container = arg1
......@@ -124,41 +123,71 @@ def _format_binary_op_str(op_str: str,
return op_str.format(arg1, arg2)
class NumpyObjectArrayMetaclass(type):
def __instancecheck__(cls, instance: Any) -> bool:
return isinstance(instance, np.ndarray) and instance.dtype == object
class NumpyObjectArray(metaclass=NumpyObjectArrayMetaclass):
pass
class ComplainingNumpyNonObjectArrayMetaclass(type):
def __instancecheck__(cls, instance: Any) -> bool:
if isinstance(instance, np.ndarray) and instance.dtype != object:
# Example usage site:
# https://github.com/illinois-ceesd/mirgecom/blob/f5d0d97c41e8c8a05546b1d1a6a2979ec8ea3554/mirgecom/inviscid.py#L148-L149
# where normal is passed in by test_lfr_flux as a 'custom-made'
# numpy array of dtype float64.
warn(
"Broadcasting container against non-object numpy array. "
"This was never documented to work and will now stop working in "
"2025. Convert the array to an object array to preserve the "
"current semantics.", DeprecationWarning, stacklevel=3)
return True
else:
return False
class ComplainingNumpyNonObjectArray(metaclass=ComplainingNumpyNonObjectArrayMetaclass):
pass
def with_container_arithmetic(
*,
bcast_number: bool = True,
_bcast_actx_array_type: Optional[bool] = None,
bcast_obj_array: Optional[bool] = None,
bcast_numpy_array: bool = False,
bcast_container_types: Optional[Tuple[type, ...]] = None,
arithmetic: bool = True,
matmul: bool = False,
bitwise: bool = False,
shift: bool = False,
_cls_has_array_context_attr: bool = False,
eq_comparison: Optional[bool] = None,
rel_comparison: Optional[bool] = None) -> Callable[[type], type]:
*,
number_bcasts_across: bool | None = None,
bcasts_across_obj_array: bool | None = None,
container_types_bcast_across: tuple[type, ...] | None = None,
arithmetic: bool = True,
matmul: bool = False,
bitwise: bool = False,
shift: bool = False,
_cls_has_array_context_attr: bool | None = None,
eq_comparison: bool | None = None,
rel_comparison: bool | None = None,
# deprecated:
bcast_number: bool | None = None,
bcast_obj_array: bool | None = None,
bcast_numpy_array: bool = False,
_bcast_actx_array_type: bool | None = None,
bcast_container_types: tuple[type, ...] | None = None,
) -> Callable[[type], type]:
"""A class decorator that implements built-in operators for array containers
by propagating the operations to the elements of the container.
:arg bcast_number: If *True*, numbers broadcast over the container
:arg number_bcasts_across: If *True*, numbers broadcast over the container
(with the container as the 'outer' structure).
:arg _bcast_actx_array_type: If *True*, instances of base array types of the
container's array context are broadcasted over the container. Can be
*True* only if the container has *_cls_has_array_context_attr* set.
Defaulted to *bcast_number* if *_cls_has_array_context_attr* is set,
else *False*.
:arg bcast_obj_array: If *True*, :mod:`numpy` object arrays broadcast over
the container. (with the container as the 'inner' structure)
:arg bcast_numpy_array: If *True*, any :class:`numpy.ndarray` will broadcast
over the container. (with the container as the 'inner' structure)
If this is set to *True*, *bcast_obj_array* must also be *True*.
:arg bcast_container_types: A sequence of container types that will broadcast
over this container (with this container as the 'outer' structure).
:arg bcasts_across_obj_array: If *True*, this container will be broadcast
across :mod:`numpy` object arrays
(with the object array as the 'outer' structure).
Add :class:`numpy.ndarray` to *container_types_bcast_across* to achieve
the 'reverse' broadcasting.
:arg container_types_bcast_across: A sequence of container types that will broadcast
across this container, with this container as the 'outer' structure.
:class:`numpy.ndarray` is permitted to be part of this sequence to
indicate that, in such broadcasting situations, this container should
be the 'outer' structure. In this case, *bcast_obj_array*
(and consequently *bcast_numpy_array*) must be *False*.
indicate that object arrays (and *only* object arrays) will be broadcast.
In this case, *bcasts_across_obj_array* must be *False*.
:arg arithmetic: Implement the conventional arithmetic operators, including
``**``, :func:`divmod`, and ``//``. Also includes ``+`` and ``-`` as well as
:func:`abs`.
......@@ -172,6 +201,8 @@ def with_container_arithmetic(
class has an ``array_context`` attribute. If so, and if :data:`__debug__`
is *True*, an additional check is performed in binary operators
to ensure that both containers use the same array context.
If *None* (the default), this value is set based on whether the class
has an ``array_context`` attribute.
Consider this argument an unstable interface. It may disappear at any moment.
Each operator class also includes the "reverse" operators if applicable.
......@@ -198,13 +229,18 @@ def with_container_arithmetic(
should nest "outside" :func:dataclass_array_container`.
"""
# {{{ handle inputs
if bcast_obj_array is None:
raise TypeError("bcast_obj_array must be specified")
# Hard-won design lessons:
#
# - Anything that special-cases np.ndarray by type is broken by design because:
# - np.ndarray is an array context array.
# - numpy object arrays can be array containers.
# Using NumpyObjectArray and NumpyNonObjectArray *may* be better?
# They're new, so there is no operational experience with them.
#
# - Broadcast rules are hard to change once established, particularly
# because one cannot grep for their use.
if rel_comparison is None:
raise TypeError("rel_comparison must be specified")
# {{{ handle inputs
if rel_comparison and eq_comparison is None:
eq_comparison = True
......@@ -212,37 +248,104 @@ def with_container_arithmetic(
if eq_comparison is None:
raise TypeError("eq_comparison must be specified")
if not bcast_obj_array and bcast_numpy_array:
raise TypeError("bcast_obj_array must be set if bcast_numpy_array is")
# {{{ handle bcast_number
if _bcast_actx_array_type is None:
if _cls_has_array_context_attr:
_bcast_actx_array_type = bcast_number
else:
_bcast_actx_array_type = False
if bcast_number is not None:
if number_bcasts_across is not None:
raise TypeError(
"may specify at most one of 'bcast_number' and "
"'number_bcasts_across'")
warn("'bcast_number' is deprecated and will be unsupported from 2025. "
"Use 'number_bcasts_across', with equivalent meaning.",
DeprecationWarning, stacklevel=2)
number_bcasts_across = bcast_number
else:
if number_bcasts_across is None:
number_bcasts_across = True
del bcast_number
# }}}
# {{{ handle bcast_obj_array
if bcast_obj_array is not None:
if bcasts_across_obj_array is not None:
raise TypeError(
"may specify at most one of 'bcast_obj_array' and "
"'bcasts_across_obj_array'")
warn("'bcast_obj_array' is deprecated and will be unsupported from 2025. "
"Use 'bcasts_across_obj_array', with equivalent meaning.",
DeprecationWarning, stacklevel=2)
bcasts_across_obj_array = bcast_obj_array
else:
if _bcast_actx_array_type and not _cls_has_array_context_attr:
raise TypeError("_bcast_actx_array_type can be True only if "
"_cls_has_array_context_attr is set.")
if bcasts_across_obj_array is None:
raise TypeError("bcasts_across_obj_array must be specified")
del bcast_obj_array
# }}}
# {{{ handle bcast_container_types
if bcast_container_types is not None:
if container_types_bcast_across is not None:
raise TypeError(
"may specify at most one of 'bcast_container_types' and "
"'container_types_bcast_across'")
warn("'bcast_container_types' is deprecated and will be unsupported from 2025. "
"Use 'container_types_bcast_across', with equivalent meaning.",
DeprecationWarning, stacklevel=2)
container_types_bcast_across = bcast_container_types
else:
if container_types_bcast_across is None:
container_types_bcast_across = ()
del bcast_container_types
# }}}
if rel_comparison is None:
raise TypeError("rel_comparison must be specified")
if bcast_numpy_array:
warn("'bcast_numpy_array=True' is deprecated and will be unsupported"
" from 2025.", DeprecationWarning, stacklevel=2)
if _bcast_actx_array_type:
raise ValueError("'bcast_numpy_array' and '_bcast_actx_array_type'"
" cannot be both set.")
if not bcasts_across_obj_array and bcast_numpy_array:
raise TypeError("bcast_obj_array must be set if bcast_numpy_array is")
if bcast_numpy_array:
def numpy_pred(name: str) -> str:
return f"isinstance({name}, np.ndarray)"
elif bcast_obj_array:
return f"is_numpy_array({name})"
elif bcasts_across_obj_array:
def numpy_pred(name: str) -> str:
return f"isinstance({name}, np.ndarray) and {name}.dtype.char == 'O'"
else:
def numpy_pred(name: str) -> str:
return "False" # optimized away
if bcast_container_types is None:
bcast_container_types = ()
bcast_container_types_count = len(bcast_container_types)
if np.ndarray in bcast_container_types and bcast_obj_array:
if np.ndarray in container_types_bcast_across and bcasts_across_obj_array:
raise ValueError("If numpy.ndarray is part of bcast_container_types, "
"bcast_obj_array must be False.")
numpy_check_types: list[type] = [NumpyObjectArray, ComplainingNumpyNonObjectArray]
container_types_bcast_across = tuple(
new_ct
for old_ct in container_types_bcast_across
for new_ct in
(numpy_check_types
if old_ct is np.ndarray
else [old_ct])
)
desired_op_classes = set()
if arithmetic:
desired_op_classes.add(_OpClass.ARITHMETIC)
......@@ -260,6 +363,64 @@ def with_container_arithmetic(
# }}}
def wrap(cls: Any) -> Any:
if not hasattr(cls, "__array_ufunc__"):
warn(f"{cls} does not have __array_ufunc__ set. "
"This will cause numpy to attempt broadcasting, in a way that "
"is likely undesired. "
f"To avoid this, set __array_ufunc__ = None in {cls}.",
stacklevel=2)
cls_has_array_context_attr: bool | None = _cls_has_array_context_attr
bcast_actx_array_type: bool | None = _bcast_actx_array_type
if cls_has_array_context_attr is None and hasattr(cls, "array_context"):
raise TypeError(
f"{cls} has an 'array_context' attribute, but it does not "
"set '_cls_has_array_context_attr' to *True* when calling "
"with_container_arithmetic. This is being interpreted "
"as '.array_context' being permitted to fail "
"with an exception, which is no longer allowed. "
f"If {cls.__name__}.array_context will not fail, pass "
"'_cls_has_array_context_attr=True'. "
"If you do not want container arithmetic to make "
"use of the array context, set "
"'_cls_has_array_context_attr=False'.")
if bcast_actx_array_type is None:
if cls_has_array_context_attr:
if number_bcasts_across:
bcast_actx_array_type = cls_has_array_context_attr
else:
bcast_actx_array_type = False
else:
if bcast_actx_array_type and not cls_has_array_context_attr:
raise TypeError("_bcast_actx_array_type can be True only if "
"_cls_has_array_context_attr is set.")
if bcast_actx_array_type:
if _bcast_actx_array_type:
warn(
f"Broadcasting array context array types across {cls} "
"has been explicitly "
"enabled. As of 2025, this will stop working. "
"There is no replacement as of right now. "
"See the discussion in "
"https://github.com/inducer/arraycontext/pull/190. "
"To opt out now (and avoid this warning), "
"pass _bcast_actx_array_type=False. ",
DeprecationWarning, stacklevel=2)
else:
warn(
f"Broadcasting array context array types across {cls} "
"has been implicitly "
"enabled. As of 2025, this will no longer work. "
"There is no replacement as of right now. "
"See the discussion in "
"https://github.com/inducer/arraycontext/pull/190. "
"To opt out now (and avoid this warning), "
"pass _bcast_actx_array_type=False.",
DeprecationWarning, stacklevel=2)
if (not hasattr(cls, "_serialize_init_arrays_code")
or not hasattr(cls, "_deserialize_init_arrays_code")):
raise TypeError(f"class '{cls.__name__}' must provide serialization "
......@@ -270,38 +431,66 @@ def with_container_arithmetic(
from pytools.codegen import CodeGenerator, Indentation
gen = CodeGenerator()
gen("""
gen(f"""
from numbers import Number
import numpy as np
from arraycontext import ArrayContainer
from warnings import warn
def _raise_if_actx_none(actx):
if actx is None:
raise ValueError("array containers with frozen arrays "
"cannot be operated upon")
return actx
def is_numpy_array(arg):
if isinstance(arg, np.ndarray):
if arg.dtype != "O":
warn("Operand is a non-object numpy array, "
"and the broadcasting behavior of this array container "
"({cls}) "
"is influenced by this because of its use of "
"the deprecated bcast_numpy_array. This broadcasting "
"behavior will change in 2025. If you would like the "
"broadcasting behavior to stay the same, make sure "
"to convert the passed numpy array to an "
"object array.",
DeprecationWarning, stacklevel=3)
return True
else:
return False
""")
gen("")
if bcast_container_types:
for i, bct in enumerate(bcast_container_types):
if container_types_bcast_across:
for i, bct in enumerate(container_types_bcast_across):
gen(f"from {bct.__module__} import {bct.__qualname__} as _bctype{i}")
gen("")
outer_bcast_type_names = tuple([
f"_bctype{i}" for i in range(bcast_container_types_count)
])
if bcast_number:
outer_bcast_type_names += ("Number",)
container_type_names_bcast_across = tuple(
f"_bctype{i}" for i in range(len(container_types_bcast_across)))
if number_bcasts_across:
container_type_names_bcast_across += ("Number",)
def same_key(k1: T, k2: T) -> T:
assert k1 == k2
return k1
def tup_str(t: Tuple[str, ...]) -> str:
def tup_str(t: tuple[str, ...]) -> str:
if not t:
return "()"
else:
return "(%s,)" % ", ".join(t)
return "({},)".format(", ".join(t))
gen(f"cls._outer_bcast_types = {tup_str(container_type_names_bcast_across)}")
gen("cls._container_types_bcast_across = "
f"{tup_str(container_type_names_bcast_across)}")
gen(f"cls._bcast_numpy_array = {bcast_numpy_array}")
gen(f"cls._bcast_obj_array = {bcasts_across_obj_array}")
gen(f"cls._bcasts_across_obj_array = {bcasts_across_obj_array}")
gen("")
# {{{ unary operators
......@@ -344,35 +533,43 @@ def with_container_arithmetic(
continue
# {{{ "forward" binary operators
zip_init_args = cls._deserialize_init_arrays_code("arg1", {
same_key(key_arg1, key_arg2):
_format_binary_op_str(op_str, expr_arg1, expr_arg2)
for (key_arg1, expr_arg1), (key_arg2, expr_arg2) in zip(
cls._serialize_init_arrays_code("arg1").items(),
cls._serialize_init_arrays_code("arg2").items())
cls._serialize_init_arrays_code("arg2").items(),
strict=True)
})
bcast_same_cls_init_args = cls._deserialize_init_arrays_code("arg1", {
bcast_init_args_arg1_is_outer = cls._deserialize_init_arrays_code("arg1", {
key_arg1: _format_binary_op_str(op_str, expr_arg1, "arg2")
for key_arg1, expr_arg1 in
cls._serialize_init_arrays_code("arg1").items()
})
bcast_init_args_arg2_is_outer = cls._deserialize_init_arrays_code("arg2", {
key_arg2: _format_binary_op_str(op_str, "arg1", expr_arg2)
for key_arg2, expr_arg2 in
cls._serialize_init_arrays_code("arg2").items()
})
# {{{ "forward" binary operators
gen(f"def {fname}(arg1, arg2):")
with Indentation(gen):
gen("if arg2.__class__ is cls:")
with Indentation(gen):
if __debug__ and _cls_has_array_context_attr:
if __debug__ and cls_has_array_context_attr:
gen("""
if arg1.array_context is not arg2.array_context:
arg1_actx = arg1.array_context
arg2_actx = arg2.array_context
if arg1_actx is not arg2_actx:
msg = ("array contexts of both arguments "
"must match")
if arg1.array_context is None:
if arg1_actx is None:
raise ValueError(msg
+ ": left operand is frozen "
"(i.e. has no array context)")
elif arg2.array_context is None:
elif arg2_actx is None:
raise ValueError(msg
+ ": right operand is frozen "
"(i.e. has no array context)")
......@@ -380,26 +577,41 @@ def with_container_arithmetic(
raise ValueError(msg)""")
gen(f"return cls({zip_init_args})")
if _bcast_actx_array_type:
if bcast_actx_array_type:
if __debug__:
bcast_actx_ary_types: Tuple[str, ...] = (
"*_raise_if_actx_none(arg1.array_context).array_types",)
bcast_actx_ary_types: tuple[str, ...] = (
"*_raise_if_actx_none("
"arg1.array_context).array_types",)
else:
bcast_actx_ary_types = ("*arg1.array_context.array_types",)
bcast_actx_ary_types = (
"*arg1.array_context.array_types",)
else:
bcast_actx_ary_types = ()
gen(f"""
if {bool(outer_bcast_type_names)}: # optimized away
if isinstance(arg2,
{tup_str(outer_bcast_type_names
+ bcast_actx_ary_types)}):
return cls({bcast_same_cls_init_args})
if {numpy_pred("arg2")}:
result = np.empty_like(arg2, dtype=object)
for i in np.ndindex(arg2.shape):
result[i] = {op_str.format("arg1", "arg2[i]")}
return result
if {bool(container_type_names_bcast_across)}: # optimized away
if isinstance(arg2,
{tup_str(container_type_names_bcast_across
+ bcast_actx_ary_types)}):
if __debug__:
if isinstance(arg2, {tup_str(bcast_actx_ary_types)}):
warn("Broadcasting {cls} over array "
f"context array type {{type(arg2)}} is deprecated "
"and will no longer work in 2025. "
"There is no replacement as of right now. "
"See the discussion in "
"https://github.com/inducer/arraycontext/"
"pull/190. ",
DeprecationWarning, stacklevel=2)
return cls({bcast_init_args_arg1_is_outer})
return NotImplemented
""")
gen(f"cls.__{dunder_name}__ = {fname}")
......@@ -411,19 +623,15 @@ def with_container_arithmetic(
if reversible:
fname = f"_{cls.__name__.lower()}_r{dunder_name}"
bcast_init_args = cls._deserialize_init_arrays_code("arg2", {
key_arg2: _format_binary_op_str(
op_str, "arg1", expr_arg2)
for key_arg2, expr_arg2 in
cls._serialize_init_arrays_code("arg2").items()
})
if _bcast_actx_array_type:
if bcast_actx_array_type:
if __debug__:
bcast_actx_ary_types = (
"*_raise_if_actx_none(arg2.array_context).array_types",)
"*_raise_if_actx_none("
"arg2.array_context).array_types",)
else:
bcast_actx_ary_types = ("*arg2.array_context.array_types",)
bcast_actx_ary_types = (
"*arg2.array_context.array_types",)
else:
bcast_actx_ary_types = ()
......@@ -431,16 +639,30 @@ def with_container_arithmetic(
def {fname}(arg2, arg1):
# assert other.__cls__ is not cls
if {bool(outer_bcast_type_names)}: # optimized away
if isinstance(arg1,
{tup_str(outer_bcast_type_names
+ bcast_actx_ary_types)}):
return cls({bcast_init_args})
if {numpy_pred("arg1")}:
result = np.empty_like(arg1, dtype=object)
for i in np.ndindex(arg1.shape):
result[i] = {op_str.format("arg1[i]", "arg2")}
return result
if {bool(container_type_names_bcast_across)}: # optimized away
if isinstance(arg1,
{tup_str(container_type_names_bcast_across
+ bcast_actx_ary_types)}):
if __debug__:
if isinstance(arg1,
{tup_str(bcast_actx_ary_types)}):
warn("Broadcasting {cls} over array "
f"context array type {{type(arg1)}} "
"is deprecated "
"and will no longer work in 2025."
"There is no replacement as of right now. "
"See the discussion in "
"https://github.com/inducer/arraycontext/"
"pull/190. ",
DeprecationWarning, stacklevel=2)
return cls({bcast_init_args_arg2_is_outer})
return NotImplemented
cls.__r{dunder_name}__ = {fname}""")
......
arraycontext/container/dataclass.py
View file @ 21a89f1b
......@@ -4,6 +4,7 @@
.. currentmodule:: arraycontext
.. autofunction:: dataclass_array_container
"""
from __future__ import annotations
__copyright__ = """
......@@ -30,34 +31,170 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from dataclasses import fields
from collections.abc import Mapping, Sequence
from dataclasses import fields, is_dataclass
from typing import NamedTuple, Union, get_args, get_origin
from arraycontext.container import is_array_container_type
# {{{ dataclass containers
class _Field(NamedTuple):
"""Small lookalike for :class:`dataclasses.Field`."""
init: bool
name: str
type: type
def is_array_type(tp: type) -> bool:
from arraycontext import Array
return tp is Array or is_array_container_type(tp)
def dataclass_array_container(cls: type) -> type:
"""A class decorator that makes the class to which it is applied a
"""A class decorator that makes the class to which it is applied an
:class:`ArrayContainer` by registering appropriate implementations of
:func:`serialize_container` and :func:`deserialize_container`.
*cls* must be a :func:`~dataclasses.dataclass`.
Attributes that are not array containers are allowed. In order to decide
whether an attribute is an array container, the declared attribute type
is checked by the criteria from :func:`is_array_container`.
is checked by the criteria from :func:`is_array_container_type`. This
includes some support for type annotations:
* a :class:`typing.Union` of array containers is considered an array container.
* other type annotations, e.g. :class:`typing.Optional`, are not considered
array containers, even if they wrap one.
.. note::
When type annotations are strings (e.g. because of
``from __future__ import annotations``),
this function relies on :func:`inspect.get_annotations`
(with ``eval_str=True``) to obtain type annotations. This
means that *cls* must live in a module that is importable.
"""
from dataclasses import is_dataclass
from types import GenericAlias, UnionType
assert is_dataclass(cls)
array_fields = [
f for f in fields(cls) if is_array_container_type(f.type)]
non_array_fields = [
f for f in fields(cls) if not is_array_container_type(f.type)]
def is_array_field(f: _Field) -> bool:
field_type = f.type
# NOTE: unions of array containers are treated separately to handle
# unions of only array containers, e.g. `Union[np.ndarray, Array]`, as
# they can work seamlessly with arithmetic and traversal.
#
# `Optional[ArrayContainer]` is not allowed, since `None` is not
# handled by `with_container_arithmetic`, which is the common case
# for current container usage. Other type annotations, e.g.
# `Tuple[Container, Container]`, are also not allowed, as they do not
# work with `with_container_arithmetic`.
#
# This is not set in stone, but mostly driven by current usage!
origin = get_origin(field_type)
# NOTE: `UnionType` is returned when using `Type1 | Type2`
if origin in (Union, UnionType):
if all(is_array_type(arg) for arg in get_args(field_type)):
return True
else:
raise TypeError(
f"Field '{f.name}' union contains non-array container "
"arguments. All arguments must be array containers.")
# NOTE: this should never happen due to using `inspect.get_annotations`
assert not isinstance(field_type, str)
if __debug__:
if not f.init:
raise ValueError(
f"Field with 'init=False' not allowed: '{f.name}'")
# NOTE:
# * `GenericAlias` catches typed `list`, `tuple`, etc.
# * `_BaseGenericAlias` catches `List`, `Tuple`, etc.
# * `_SpecialForm` catches `Any`, `Literal`, etc.
from typing import ( # type: ignore[attr-defined]
_BaseGenericAlias,
_SpecialForm,
)
if isinstance(field_type, GenericAlias | _BaseGenericAlias | _SpecialForm):
# NOTE: anything except a Union is not allowed
raise TypeError(
f"Typing annotation not supported on field '{f.name}': "
f"'{field_type!r}'")
if not isinstance(field_type, type):
raise TypeError(
f"Field '{f.name}' not an instance of 'type': "
f"'{field_type!r}'")
return is_array_type(field_type)
from pytools import partition
array_fields = _get_annotated_fields(cls)
array_fields, non_array_fields = partition(is_array_field, array_fields)
if not array_fields:
raise ValueError(f"'{cls}' must have fields with array container type "
"in order to use the 'dataclass_array_container' decorator")
return _inject_dataclass_serialization(cls, array_fields, non_array_fields)
def _get_annotated_fields(cls: type) -> Sequence[_Field]:
"""Get a list of fields in the class *cls* with evaluated types.
If any of the fields in *cls* have type annotations that are strings, e.g.
from using ``from __future__ import annotations``, this function evaluates
them using :func:`inspect.get_annotations`. Note that this requires the class
to live in a module that is importable.
:return: a list of fields.
"""
from inspect import get_annotations
result = []
cls_ann: Mapping[str, type] | None = None
for field in fields(cls):
field_type_or_str = field.type
if isinstance(field_type_or_str, str):
if cls_ann is None:
cls_ann = get_annotations(cls, eval_str=True)
field_type = cls_ann[field.name]
else:
field_type = field_type_or_str
result.append(_Field(init=field.init, name=field.name, type=field_type))
return result
def _inject_dataclass_serialization(
cls: type,
array_fields: Sequence[_Field],
non_array_fields: Sequence[_Field]) -> type:
"""Implements :func:`~arraycontext.serialize_container` and
:func:`~arraycontext.deserialize_container` for the given dataclass *cls*.
This function modifies *cls* in place, so the returned value is the same
object with additional functionality.
:arg array_fields: fields of the given dataclass *cls* which are considered
array containers and should be serialized.
:arg non_array_fields: remaining fields of the dataclass *cls* which are
copied over from the template array in deserialization.
"""
assert is_dataclass(cls)
serialize_expr = ", ".join(
f"({f.name!r}, ary.{f.name})" for f in array_fields)
template_kwargs = ", ".join(
......
arraycontext/container/traversal.py
View file @ 21a89f1b
......@@ -8,9 +8,13 @@
.. autofunction:: rec_map_array_container
.. autofunction:: rec_multimap_array_container
.. autofunction:: map_reduce_array_container
.. autofunction:: multimap_reduce_array_container
.. autofunction:: rec_map_reduce_array_container
.. autofunction:: rec_multimap_reduce_array_container
.. autofunction:: stringify_array_container_tree
Traversing decorators
~~~~~~~~~~~~~~~~~~~~~
.. autofunction:: mapped_over_array_containers
......@@ -21,12 +25,27 @@ Freezing and thawing
.. autofunction:: freeze
.. autofunction:: thaw
Flattening and unflattening
~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autofunction:: flatten
.. autofunction:: unflatten
.. autofunction:: flat_size_and_dtype
Numpy conversion
~~~~~~~~~~~~~~~~
.. autofunction:: from_numpy
.. autofunction:: to_numpy
Algebraic operations
~~~~~~~~~~~~~~~~~~~~
.. autofunction:: outer
"""
from __future__ import annotations
from arraycontext.container.arithmetic import NumpyObjectArray
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
"""
......@@ -51,24 +70,39 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from typing import Any, Callable, Iterable, List, Optional, Union, Tuple
from functools import update_wrapper, partial, singledispatch
from collections.abc import Callable, Iterable
from functools import partial, singledispatch, update_wrapper
from typing import Any, cast
from warnings import warn
import numpy as np
from arraycontext.context import ArrayContext
from arraycontext.container import (
ArrayContainerT, is_array_container,
serialize_container, deserialize_container)
ArrayContainer,
NotAnArrayContainerError,
SerializationKey,
deserialize_container,
get_container_context_recursively_opt,
serialize_container,
)
from arraycontext.context import (
Array,
ArrayContext,
ArrayOrContainer,
ArrayOrContainerOrScalar,
ArrayOrContainerT,
ArrayT,
ScalarLike,
)
# {{{ array container traversal helpers
def _map_array_container_impl(
f: Callable[[Any], Any],
ary: ArrayContainerT, *,
leaf_cls: Optional[type] = None,
recursive: bool = False) -> ArrayContainerT:
f: Callable[[ArrayOrContainer], ArrayOrContainer],
ary: ArrayOrContainer, *,
leaf_cls: type | None = None,
recursive: bool = False) -> ArrayOrContainer:
"""Helper for :func:`rec_map_array_container`.
:param leaf_cls: class on which we call *f* directly. This is mostly
......@@ -76,15 +110,18 @@ def _map_array_container_impl(
specific container classes. By default, the recursion is stopped when
a non-:class:`ArrayContainer` class is encountered.
"""
def rec(_ary: ArrayContainerT) -> ArrayContainerT:
if type(_ary) is leaf_cls: # type(ary) is never None
return f(_ary)
elif is_array_container(_ary):
return deserialize_container(_ary, [
(key, frec(subary)) for key, subary in serialize_container(_ary)
])
def rec(ary_: ArrayOrContainer) -> ArrayOrContainer:
if type(ary_) is leaf_cls: # type(ary) is never None
return f(ary_)
try:
iterable = serialize_container(ary_)
except NotAnArrayContainerError:
return f(ary_)
else:
return f(_ary)
return deserialize_container(ary_, [
(key, frec(subary)) for key, subary in iterable
])
frec = rec if recursive else f
return rec(ary)
......@@ -93,9 +130,10 @@ def _map_array_container_impl(
def _multimap_array_container_impl(
f: Callable[..., Any],
*args: Any,
reduce_func: Callable[[Any, Iterable[Tuple[Any, Any]]], Any] = None,
leaf_cls: Optional[type] = None,
recursive: bool = False) -> ArrayContainerT:
reduce_func: (
Callable[[ArrayContainer, Iterable[tuple[Any, Any]]], Any] | None) = None,
leaf_cls: type | None = None,
recursive: bool = False) -> ArrayOrContainer:
"""Helper for :func:`rec_multimap_array_container`.
:param leaf_cls: class on which we call *f* directly. This is mostly
......@@ -103,24 +141,33 @@ def _multimap_array_container_impl(
specific container classes. By default, the recursion is stopped when
a non-:class:`ArrayContainer` class is encountered.
"""
def rec(*_args: Any) -> Any:
template_ary = _args[container_indices[0]]
if (type(template_ary) is leaf_cls
or not is_array_container(template_ary)):
return f(*_args)
# {{{ recursive traversal
def rec(*args_: Any) -> Any:
template_ary = args_[container_indices[0]]
if type(template_ary) is leaf_cls:
return f(*args_)
try:
iterable_template = serialize_container(template_ary)
except NotAnArrayContainerError:
return f(*args_)
assert all(
type(_args[i]) is type(template_ary) for i in container_indices[1:]
type(args_[i]) is type(template_ary) for i in container_indices[1:]
), f"expected type '{type(template_ary).__name__}'"
result = []
new_args = list(_args)
new_args = list(args_)
for subarys in zip(*[
serialize_container(_args[i]) for i in container_indices
]):
for subarys in zip(
iterable_template,
*[serialize_container(args_[i]) for i in container_indices[1:]],
strict=True
):
key = None
for i, (subkey, subary) in zip(container_indices, subarys):
for i, (subkey, subary) in zip(container_indices, subarys, strict=True):
if key is None:
key = subkey
else:
......@@ -128,34 +175,63 @@ def _multimap_array_container_impl(
new_args[i] = subary
result.append((key, frec(*new_args))) # type: ignore[operator]
result.append((key, frec(*new_args)))
return process_container(template_ary, result)
# }}}
# {{{ find all containers in the argument list
container_indices: list[int] = []
for i, arg in enumerate(args):
if type(arg) is leaf_cls:
continue
try:
# FIXME: this will serialize again once `rec` is called, which is
# not great, but it doesn't seem like there's a good way to avoid it
_ = serialize_container(arg)
except NotAnArrayContainerError:
pass
else:
container_indices.append(i)
return process_container(template_ary, result) # type: ignore[operator]
# }}}
container_indices: List[int] = [
i for i, arg in enumerate(args)
if is_array_container(arg) and type(arg) is not leaf_cls]
# {{{ #containers == 0 => call `f`
if not container_indices:
return f(*args)
# }}}
# {{{ #containers == 1 => call `map_array_container`
if len(container_indices) == 1 and reduce_func is None:
# NOTE: if we just have one ArrayContainer in args, passing it through
# _map_array_container_impl should be faster
def wrapper(ary: ArrayContainerT) -> ArrayContainerT:
def wrapper(ary: ArrayOrContainerT) -> ArrayOrContainerT:
new_args = list(args)
new_args[container_indices[0]] = ary
return f(*new_args)
update_wrapper(wrapper, f)
template_ary: ArrayContainerT = args[container_indices[0]]
template_ary: ArrayContainer = args[container_indices[0]]
return _map_array_container_impl(
wrapper, template_ary,
leaf_cls=leaf_cls, recursive=recursive)
# }}}
# {{{ #containers > 1 => call `rec`
process_container = deserialize_container if reduce_func is None else reduce_func
frec = rec if recursive else f
# }}}
return rec(*args)
# }}}
......@@ -163,9 +239,33 @@ def _multimap_array_container_impl(
# {{{ array container traversal
def stringify_array_container_tree(ary: ArrayOrContainer) -> str:
"""
:returns: a string for an ASCII tree representation of the array container,
similar to `asciitree <https://github.com/mbr/asciitree>`__.
"""
def rec(lines: list[str], ary_: ArrayOrContainerT, level: int) -> None:
try:
iterable = serialize_container(ary_)
except NotAnArrayContainerError:
pass
else:
for key, subary in iterable:
key = f"{key} ({type(subary).__name__})"
indent = "" if level == 0 else f" | {' ' * 4 * (level - 1)}"
lines.append(f"{indent} +-- {key}")
rec(lines, subary, level + 1)
lines = [f"root ({type(ary).__name__})"]
rec(lines, ary, 0)
return "\n".join(lines)
def map_array_container(
f: Callable[[Any], Any],
ary: ArrayContainerT) -> ArrayContainerT:
ary: ArrayOrContainer) -> ArrayOrContainer:
r"""Applies *f* to all components of an :class:`ArrayContainer`.
Works similarly to :func:`~pytools.obj_array.obj_array_vectorize`, but
......@@ -177,12 +277,13 @@ def map_array_container(
or an instance of a base array type.
"""
try:
ser_ctr = serialize_container(ary)
except TypeError:
iterable = serialize_container(ary)
except NotAnArrayContainerError:
return f(ary)
else:
return deserialize_container(ary, [
(key, f(subary)) for key, subary in ser_ctr])
(key, f(subary)) for key, subary in iterable
])
def multimap_array_container(f: Callable[..., Any], *args: Any) -> Any:
......@@ -202,7 +303,8 @@ def multimap_array_container(f: Callable[..., Any], *args: Any) -> Any:
def rec_map_array_container(
f: Callable[[Any], Any],
ary: ArrayContainerT) -> ArrayContainerT:
ary: ArrayOrContainer,
leaf_class: type | None = None) -> ArrayOrContainer:
r"""Applies *f* recursively to an :class:`ArrayContainer`.
For a non-recursive version see :func:`map_array_container`.
......@@ -210,18 +312,32 @@ def rec_map_array_container(
:param ary: a (potentially nested) structure of :class:`ArrayContainer`\ s,
or an instance of a base array type.
"""
return _map_array_container_impl(f, ary, recursive=True)
return _map_array_container_impl(f, ary, leaf_cls=leaf_class, recursive=True)
def mapped_over_array_containers(
f: Callable[[Any], Any]) -> Callable[[ArrayContainerT], ArrayContainerT]:
f: Callable[[ArrayOrContainer], ArrayOrContainer] | None = None,
leaf_class: type | None = None) -> (
Callable[[ArrayOrContainer], ArrayOrContainer]
| Callable[
[Callable[[Any], Any]],
Callable[[ArrayOrContainer], ArrayOrContainer]]):
"""Decorator around :func:`rec_map_array_container`."""
wrapper = partial(rec_map_array_container, f)
update_wrapper(wrapper, f)
return wrapper
def decorator(g: Callable[[ArrayOrContainer], ArrayOrContainer]) -> Callable[
[ArrayOrContainer], ArrayOrContainer]:
wrapper = partial(rec_map_array_container, g, leaf_class=leaf_class)
update_wrapper(wrapper, g)
return wrapper
if f is not None:
return decorator(f)
else:
return decorator
def rec_multimap_array_container(f: Callable[..., Any], *args: Any) -> Any:
def rec_multimap_array_container(
f: Callable[..., Any],
*args: Any,
leaf_class: type | None = None) -> Any:
r"""Applies *f* recursively to multiple :class:`ArrayContainer`\ s.
For a non-recursive version see :func:`multimap_array_container`.
......@@ -229,27 +345,39 @@ def rec_multimap_array_container(f: Callable[..., Any], *args: Any) -> Any:
:param args: all :class:`ArrayContainer` arguments must be of the same
type and with the same structure (same number of components, etc.).
"""
return _multimap_array_container_impl(f, *args, recursive=True)
return _multimap_array_container_impl(
f, *args, leaf_cls=leaf_class, recursive=True)
def multimapped_over_array_containers(
f: Callable[..., Any]) -> Callable[..., Any]:
f: Callable[..., Any] | None = None,
leaf_class: type | None = None) -> (
Callable[..., Any]
| Callable[[Callable[..., Any]], Callable[..., Any]]):
"""Decorator around :func:`rec_multimap_array_container`."""
# can't use functools.partial, because its result is insufficiently
# function-y to be used as a method definition.
def wrapper(*args: Any) -> Any:
return rec_multimap_array_container(f, *args)
def decorator(g: Callable[..., Any]) -> Callable[..., Any]:
# can't use functools.partial, because its result is insufficiently
# function-y to be used as a method definition.
def wrapper(*args: Any) -> Any:
return rec_multimap_array_container(g, *args, leaf_class=leaf_class)
update_wrapper(wrapper, g)
return wrapper
if f is not None:
return decorator(f)
else:
return decorator
update_wrapper(wrapper, f)
return wrapper
# }}}
# {{{ keyed array container traversal
def keyed_map_array_container(f: Callable[[Any, Any], Any],
ary: ArrayContainerT) -> ArrayContainerT:
def keyed_map_array_container(
f: Callable[
[SerializationKey, ArrayOrContainer],
ArrayOrContainer],
ary: ArrayOrContainer) -> ArrayOrContainer:
r"""Applies *f* to all components of an :class:`ArrayContainer`.
Works similarly to :func:`map_array_container`, but *f* also takes an
......@@ -260,16 +388,20 @@ def keyed_map_array_container(f: Callable[[Any, Any], Any],
:param ary: a (potentially nested) structure of :class:`ArrayContainer`\ s,
or an instance of a base array type.
"""
if is_array_container(ary):
return deserialize_container(ary, [
(key, f(key, subary)) for key, subary in serialize_container(ary)
])
try:
iterable = serialize_container(ary)
except NotAnArrayContainerError as err:
raise ValueError(
f"Non-array container type has no key: {type(ary).__name__}") from err
else:
raise ValueError("Not an array-container, i.e. unknown key to pass.")
return deserialize_container(ary, [
(key, f(key, subary)) for key, subary in iterable
])
def rec_keyed_map_array_container(f: Callable[[Tuple[Any, ...], Any], Any],
ary: ArrayContainerT) -> ArrayContainerT:
def rec_keyed_map_array_container(
f: Callable[[tuple[SerializationKey, ...], ArrayT], ArrayT],
ary: ArrayOrContainer) -> ArrayOrContainer:
"""
Works similarly to :func:`rec_map_array_container`, except that *f* also
takes in a traversal path to the leaf array. The traversal path argument is
......@@ -277,15 +409,16 @@ def rec_keyed_map_array_container(f: Callable[[Tuple[Any, ...], Any], Any],
the current array.
"""
def rec(keys: Tuple[Union[str, int], ...],
_ary: ArrayContainerT) -> ArrayContainerT:
if is_array_container(_ary):
return deserialize_container(_ary, [
(key, rec(keys + (key,), subary))
for key, subary in serialize_container(_ary)
])
def rec(keys: tuple[SerializationKey, ...],
ary_: ArrayOrContainerT) -> ArrayOrContainerT:
try:
iterable = serialize_container(ary_)
except NotAnArrayContainerError:
return cast(ArrayOrContainerT, f(keys, cast(ArrayT, ary_)))
else:
return f(keys, _ary)
return deserialize_container(ary_, [
(key, rec((*keys, key), subary)) for key, subary in iterable
])
return rec((), ary)
......@@ -294,26 +427,108 @@ def rec_keyed_map_array_container(f: Callable[[Tuple[Any, ...], Any], Any],
# {{{ array container reductions
def map_reduce_array_container(
reduce_func: Callable[[Iterable[Any]], Any],
map_func: Callable[[Any], Any],
ary: ArrayOrContainerT) -> Array:
"""Perform a map-reduce over array containers.
:param reduce_func: callable used to reduce over the components of *ary*
if *ary* is an :class:`~arraycontext.ArrayContainer`. The callable
should be associative, as for :func:`rec_map_reduce_array_container`.
:param map_func: callable used to map a single array of type
:class:`arraycontext.ArrayContext.array_types`. Returns an array of the
same type or a scalar.
"""
try:
iterable = serialize_container(ary)
except NotAnArrayContainerError:
return map_func(ary)
else:
return reduce_func([
map_func(subary) for _, subary in iterable
])
def multimap_reduce_array_container(
reduce_func: Callable[[Iterable[Any]], Any],
map_func: Callable[..., Any],
*args: Any) -> ArrayOrContainer:
r"""Perform a map-reduce over multiple array containers.
:param reduce_func: callable used to reduce over the components of any
:class:`~arraycontext.ArrayContainer`\ s in *\*args*. The callable
should be associative, as for :func:`rec_map_reduce_array_container`.
:param map_func: callable used to map a single array of type
:class:`arraycontext.ArrayContext.array_types`. Returns an array of the
same type or a scalar.
"""
# NOTE: this wrapper matches the signature of `deserialize_container`
# to make plugging into `_multimap_array_container_impl` easier
def _reduce_wrapper(
ary: ArrayContainer, iterable: Iterable[tuple[Any, Any]]
) -> Array:
return reduce_func([subary for _, subary in iterable])
return _multimap_array_container_impl(
map_func, *args,
reduce_func=_reduce_wrapper, leaf_cls=None, recursive=False)
def rec_map_reduce_array_container(
reduce_func: Callable[[Iterable[Any]], Any],
map_func: Callable[[Any], Any],
ary: ArrayContainerT) -> Any:
ary: ArrayOrContainer,
leaf_class: type | None = None) -> ArrayOrContainer:
"""Perform a map-reduce over array containers recursively.
:param reduce_func: callable used to reduce over the components of the
:class:`~arraycontext.ArrayContainer`.
:param map_func: callable used to map a single component of the
:class:`~arraycontext.ArrayContainer`. The callable takes arrays of
type :class:`arraycontext.ArrayContext.array_types` and returns an
array of the same type or a scalar.
:param reduce_func: callable used to reduce over the components of *ary*
(and those of its sub-containers) if *ary* is a
:class:`~arraycontext.ArrayContainer`. Must be associative.
:param map_func: callable used to map a single array of type
:class:`arraycontext.ArrayContext.array_types`. Returns an array of the
same type or a scalar.
.. note::
The traversal order is unspecified. *reduce_func* must be associative in
order to guarantee a sensible result. This is because *reduce_func* may be
called on subsets of the component arrays, and then again (potentially
multiple times) on the results. As an example, consider a container made up
of two sub-containers, *subcontainer0* and *subcontainer1*, that each
contain two component arrays, *array0* and *array1*. The same result must be
computed whether traversing recursively::
reduce_func([
reduce_func([
map_func(subcontainer0.array0),
map_func(subcontainer0.array1)]),
reduce_func([
map_func(subcontainer1.array0),
map_func(subcontainer1.array1)])])
reducing all of the arrays at once::
reduce_func([
map_func(subcontainer0.array0),
map_func(subcontainer0.array1),
map_func(subcontainer1.array0),
map_func(subcontainer1.array1)])
or any other such traversal.
"""
def rec(_ary: ArrayContainerT) -> ArrayContainerT:
if is_array_container(_ary):
return reduce_func([
rec(subary) for _, subary in serialize_container(_ary)
])
def rec(ary_: ArrayOrContainerT) -> ArrayOrContainerT:
if type(ary_) is leaf_class:
return map_func(ary_)
else:
return map_func(_ary)
try:
iterable = serialize_container(ary_)
except NotAnArrayContainerError:
return map_func(ary_)
else:
return reduce_func([
rec(subary) for _, subary in iterable
])
return rec(ary)
......@@ -321,34 +536,41 @@ def rec_map_reduce_array_container(
def rec_multimap_reduce_array_container(
reduce_func: Callable[[Iterable[Any]], Any],
map_func: Callable[..., Any],
*args: Any) -> Any:
"""Perform a map-reduce over multiple array containers recursively.
:param reduce_func: callable used to reduce over the components of the
:class:`~arraycontext.ArrayContainer`.
:param map_func: callable used to map a single component of the
:class:`~arraycontext.ArrayContainer`. The callable takes arrays of
type :class:`arraycontext.ArrayContext.array_types` and returns an
array of the same type or a scalar.
*args: Any,
leaf_class: type | None = None) -> ArrayOrContainer:
r"""Perform a map-reduce over multiple array containers recursively.
:param reduce_func: callable used to reduce over the components of any
:class:`~arraycontext.ArrayContainer`\ s in *\*args* (and those of their
sub-containers). Must be associative.
:param map_func: callable used to map a single array of type
:class:`arraycontext.ArrayContext.array_types`. Returns an array of the
same type or a scalar.
.. note::
The traversal order is unspecified. *reduce_func* must be associative in
order to guarantee a sensible result. See
:func:`rec_map_reduce_array_container` for additional details.
"""
# NOTE: this wrapper matches the signature of `deserialize_container`
# to make plugging into `_multimap_array_container_impl` easier
def _reduce_wrapper(ary: Any, iterable: Iterable[Tuple[Any, Any]]) -> Any:
def _reduce_wrapper(
ary: ArrayContainer, iterable: Iterable[tuple[Any, Any]]) -> Any:
return reduce_func([subary for _, subary in iterable])
return _multimap_array_container_impl(
map_func, *args,
reduce_func=_reduce_wrapper, leaf_cls=None, recursive=True)
reduce_func=_reduce_wrapper, leaf_cls=leaf_class, recursive=True)
# }}}
# {{{ freeze/thaw
@singledispatch
def freeze(
ary: ArrayContainerT,
actx: Optional[ArrayContext] = None) -> ArrayContainerT:
ary: ArrayOrContainerT,
actx: ArrayContext | None = None) -> ArrayOrContainerT:
r"""Freezes recursively by going through all components of the
:class:`ArrayContainer` *ary*.
......@@ -359,20 +581,34 @@ def freeze(
See :meth:`ArrayContext.thaw`.
"""
if is_array_container(ary):
return map_array_container(partial(freeze, actx=actx), ary)
if actx is None:
warn("Calling freeze(ary) without specifying actx is deprecated, explicitly"
" call actx.freeze(ary) instead. This will stop working in 2023.",
DeprecationWarning, stacklevel=2)
actx = get_container_context_recursively_opt(ary)
else:
if actx is None:
raise TypeError(
f"cannot freeze arrays of type {type(ary).__name__} "
"when actx is not supplied. Try calling actx.freeze "
"directly or supplying an array context")
else:
return actx.freeze(ary)
warn("Calling freeze(ary, actx) is deprecated, call actx.freeze(ary)"
" instead. This will stop working in 2023.",
DeprecationWarning, stacklevel=2)
if __debug__:
rec_actx = get_container_context_recursively_opt(ary)
if (rec_actx is not None) and (rec_actx is not actx):
raise ValueError("Supplied array context does not agree with"
" the one obtained by traversing 'ary'.")
@singledispatch
def thaw(ary: ArrayContainerT, actx: ArrayContext) -> ArrayContainerT:
if actx is None:
raise TypeError(
f"cannot freeze arrays of type {type(ary).__name__} "
"when actx is not supplied. Try calling actx.freeze "
"directly or supplying an array context")
return actx.freeze(ary)
def thaw(ary: ArrayOrContainerT, actx: ArrayContext) -> ArrayOrContainerT:
r"""Thaws recursively by going through all components of the
:class:`ArrayContainer` *ary*.
......@@ -392,43 +628,362 @@ def thaw(ary: ArrayContainerT, actx: ArrayContext) -> ArrayContainerT:
in :mod:`meshmode`. This was necessary because
:func:`~functools.singledispatch` only dispatches on the first argument.
"""
if is_array_container(ary):
return deserialize_container(ary, [
(key, thaw(subary, actx))
for key, subary in serialize_container(ary)
])
warn("Calling thaw(ary, actx) is deprecated, call actx.thaw(ary) instead."
" This will stop working in 2023.",
DeprecationWarning, stacklevel=2)
if __debug__:
rec_actx = get_container_context_recursively_opt(ary)
if rec_actx is not None:
raise ValueError("cannot thaw a container that already has an array"
" context.")
return actx.thaw(ary)
# }}}
# {{{ with_array_context
@singledispatch
def with_array_context(ary: ArrayOrContainerT,
actx: ArrayContext | None) -> ArrayOrContainerT:
"""
Recursively associates *actx* to all the components of *ary*.
Array container types may use :func:`functools.singledispatch` ``.register``
to register container-specific implementations. See `this issue
<https://github.com/inducer/arraycontext/issues/162>`__ for discussion of
the future of this functionality.
"""
try:
iterable = serialize_container(ary)
except NotAnArrayContainerError:
return ary
else:
return actx.thaw(ary)
return deserialize_container(ary, [(key, with_array_context(subary, actx))
for key, subary in iterable])
# }}}
# {{{ flatten / unflatten
def flatten(
ary: ArrayOrContainer, actx: ArrayContext, *,
leaf_class: type | None = None,
) -> Any:
"""Convert all arrays in the :class:`~arraycontext.ArrayContainer`
into single flat array of a type :attr:`arraycontext.ArrayContext.array_types`.
The operation requires :attr:`arraycontext.ArrayContext.np` to have
``ravel`` and ``concatenate`` methods implemented. The order in which the
individual leaf arrays appear in the final array is dependent on the order
given by :func:`~arraycontext.serialize_container`.
If *leaf_class* is given, then :func:`unflatten` will not be able to recover
the original *ary*.
:arg leaf_class: an :class:`~arraycontext.ArrayContainer` class on which
the recursion is stopped (subclasses are not considered). If given, only
the entries of this type are flattened and the rest of the tree
structure is left as is. By default, the recursion is stopped when
a non-:class:`~arraycontext.ArrayContainer` is found, which results in
the whole input container *ary* being flattened.
"""
common_dtype = None
def _flatten(subary: ArrayOrContainer) -> list[Array]:
nonlocal common_dtype
try:
iterable = serialize_container(subary)
except NotAnArrayContainerError:
subary_c = cast(Array, subary)
if common_dtype is None:
common_dtype = subary_c.dtype
if subary_c.dtype != common_dtype:
raise ValueError("arrays in container have different dtypes: "
f"got {subary_c.dtype}, expected {common_dtype}") from None
try:
flat_subary = actx.np.ravel(subary_c, order="C")
except ValueError as exc:
# NOTE: we can't do much if the array context fails to ravel,
# since it is the one responsible for the actual memory layout
if hasattr(subary_c, "strides"):
strides_msg = f" and strides {subary_c.strides}"
else:
strides_msg = ""
raise NotImplementedError(
f"'{type(actx).__name__}.np.ravel' failed to reshape "
f"an array with shape {subary_c.shape}{strides_msg}. "
"This functionality needs to be implemented by the "
"array context.") from exc
result = [flat_subary]
else:
result = []
for _, isubary in iterable:
result.extend(_flatten(isubary))
return result
def _flatten_without_leaf_class(subary: ArrayOrContainer) -> Any:
result = _flatten(subary)
if len(result) == 1:
return result[0]
else:
return actx.np.concatenate(result)
def _flatten_with_leaf_class(subary: ArrayOrContainer) -> Any:
if type(subary) is leaf_class:
return _flatten_without_leaf_class(subary)
try:
iterable = serialize_container(subary)
except NotAnArrayContainerError:
return subary
else:
return deserialize_container(subary, [
(key, _flatten_with_leaf_class(isubary))
for key, isubary in iterable
])
if leaf_class is None:
return _flatten_without_leaf_class(ary)
else:
return _flatten_with_leaf_class(ary)
def unflatten(
template: ArrayOrContainerT, ary: Array,
actx: ArrayContext, *,
strict: bool = True) -> ArrayOrContainerT:
"""Unflatten an array *ary* produced by :func:`flatten` back into an
:class:`~arraycontext.ArrayContainer`.
The order and sizes of each slice into *ary* are determined by the
array container *template*.
:arg ary: a flat one-dimensional array with a size that matches the
number of entries in *template*.
:arg strict: if *True* additional :class:`~numpy.dtype` and stride
checking is performed on the unflattened array. Otherwise, these
checks are skipped.
"""
# NOTE: https://github.com/python/mypy/issues/7057
offset = 0
common_dtype = None
def _unflatten(template_subary: ArrayOrContainer) -> ArrayOrContainer:
nonlocal offset, common_dtype
try:
iterable = serialize_container(template_subary)
except NotAnArrayContainerError:
template_subary_c = cast(Array, template_subary)
# {{{ validate subary
if (offset + template_subary_c.size) > ary.size:
raise ValueError("'template' and 'ary' sizes do not match: "
"'template' is too large") from None
if strict:
if template_subary_c.dtype != ary.dtype:
raise ValueError("'template' dtype does not match 'ary': "
f"got {template_subary_c.dtype}, expected {ary.dtype}"
) from None
else:
# NOTE: still require that *template* has a uniform dtype
if common_dtype is None:
common_dtype = template_subary_c.dtype
else:
if common_dtype != template_subary_c.dtype:
raise ValueError("arrays in 'template' have different "
f"dtypes: got {template_subary_c.dtype}, but "
f"expected {common_dtype}.") from None
# }}}
# {{{ reshape
flat_subary = ary[offset:offset + template_subary_c.size]
try:
subary = actx.np.reshape(flat_subary,
template_subary_c.shape, order="C")
except ValueError as exc:
# NOTE: we can't do much if the array context fails to reshape,
# since it is the one responsible for the actual memory layout
raise NotImplementedError(
f"'{type(actx).__name__}.np.reshape' failed to reshape "
f"the flat array into shape {template_subary_c.shape}. "
"This functionality needs to be implemented by the "
"array context.") from exc
# }}}
# {{{ check strides
if strict and hasattr(template_subary_c, "strides"): # noqa: SIM102
# Checking strides for 0 sized arrays is ill-defined
# since they cannot be indexed
if (
# Mypy has a point: nobody promised a .strides attribute.
template_subary_c.strides != subary.strides
and template_subary_c.size != 0
):
raise ValueError(
# Mypy has a point: nobody promised a .strides attribute.
f"strides do not match template: got {subary.strides}, "
f"expected {template_subary_c.strides}") from None
# }}}
offset += template_subary_c.size
return subary
else:
return deserialize_container(template_subary, [
(key, _unflatten(isubary)) for key, isubary in iterable
])
if not isinstance(ary, actx.array_types):
raise TypeError("'ary' does not have a type supported by the provided "
f"array context: got '{type(ary).__name__}', expected one of "
f"{actx.array_types}")
if len(ary.shape) != 1:
raise ValueError(
"only one dimensional arrays can be unflattened: "
f"'ary' has shape {ary.shape}")
result = _unflatten(template)
if offset != ary.size:
raise ValueError("'template' and 'ary' sizes do not match: "
"'ary' is too large")
return cast(ArrayOrContainerT, result)
def flat_size_and_dtype(
ary: ArrayOrContainer) -> tuple[int, np.dtype[Any] | None]:
"""
:returns: a tuple ``(size, dtype)`` that would be the length and
:class:`numpy.dtype` of the one-dimensional array returned by
:func:`flatten`.
"""
common_dtype = None
def _flat_size(subary: ArrayOrContainer) -> int:
nonlocal common_dtype
try:
iterable = serialize_container(subary)
except NotAnArrayContainerError:
subary_c = cast(Array, subary)
if common_dtype is None:
common_dtype = subary_c.dtype
if subary_c.dtype != common_dtype:
raise ValueError("arrays in container have different dtypes: "
f"got {subary_c.dtype}, expected {common_dtype}") from None
return subary_c.size
else:
return sum(_flat_size(isubary) for _, isubary in iterable)
size = _flat_size(ary)
return size, common_dtype
# }}}
# {{{ numpy conversion
def from_numpy(ary: Any, actx: ArrayContext) -> Any:
def from_numpy(
ary: np.ndarray | ScalarLike,
actx: ArrayContext) -> ArrayOrContainerOrScalar:
"""Convert all :mod:`numpy` arrays in the :class:`~arraycontext.ArrayContainer`
to the base array type of :class:`~arraycontext.ArrayContext`.
The conversion is done using :meth:`arraycontext.ArrayContext.from_numpy`.
"""
def _from_numpy(subary: Any) -> Any:
if isinstance(subary, np.ndarray) and subary.dtype != "O":
return actx.from_numpy(subary)
elif is_array_container(subary):
return map_array_container(_from_numpy, subary)
else:
raise TypeError(f"unrecognized array type: '{type(subary).__name__}'")
warn("Calling from_numpy(ary, actx) is deprecated, call actx.from_numpy(ary)"
" instead. This will stop working in 2023.",
DeprecationWarning, stacklevel=2)
return _from_numpy(ary)
return actx.from_numpy(ary)
def to_numpy(ary: Any, actx: ArrayContext) -> Any:
def to_numpy(ary: ArrayOrContainer, actx: ArrayContext) -> ArrayOrContainer:
"""Convert all arrays in the :class:`~arraycontext.ArrayContainer` to
:mod:`numpy` using the provided :class:`~arraycontext.ArrayContext` *actx*.
The conversion is done using :meth:`arraycontext.ArrayContext.to_numpy`.
"""
return rec_map_array_container(actx.to_numpy, ary)
warn("Calling to_numpy(ary, actx) is deprecated, call actx.to_numpy(ary)"
" instead. This will stop working in 2023.",
DeprecationWarning, stacklevel=2)
return actx.to_numpy(ary)
# }}}
# {{{ algebraic operations
def outer(a: Any, b: Any) -> Any:
"""
Compute the outer product of *a* and *b* while allowing either of them
to be an :class:`ArrayContainer`.
Tweaks the behavior of :func:`numpy.outer` to return a lower-dimensional
object if either/both of *a* and *b* are scalars (whereas :func:`numpy.outer`
always returns a matrix). Here the definition of "scalar" includes
all non-array-container types and any scalar-like array container types.
If *a* and *b* are both array containers, the result will have the same type
as *a*. If both are array containers and neither is an object array, they must
have the same type.
"""
def treat_as_scalar(x: Any) -> bool:
try:
serialize_container(x)
except NotAnArrayContainerError:
return True
else:
return (
not isinstance(x, np.ndarray)
# This condition is whether "ndarrays should broadcast inside x".
and NumpyObjectArray not in x.__class__._outer_bcast_types)
a_is_ndarray = isinstance(a, np.ndarray)
b_is_ndarray = isinstance(b, np.ndarray)
if a_is_ndarray and a.dtype != object:
raise TypeError("passing a non-object numpy array is not allowed")
if b_is_ndarray and b.dtype != object:
raise TypeError("passing a non-object numpy array is not allowed")
if treat_as_scalar(a) or treat_as_scalar(b):
return a*b
elif a_is_ndarray and b_is_ndarray:
return np.outer(a, b)
elif a_is_ndarray or b_is_ndarray:
return map_array_container(lambda x: outer(x, b), a)
else:
if type(a) is not type(b):
raise TypeError(
"both arguments must have the same type if they are both "
"non-object-array array containers.")
return multimap_array_container(lambda x, y: outer(x, y), a, b)
# }}}
......
arraycontext/context.py
View file @ 21a89f1b
# mypy: disallow-untyped-defs
"""
.. _freeze-thaw:
......@@ -39,7 +41,7 @@ Here are some rules of thumb to use when dealing with thawing and freezing:
- Note that array contexts need not necessarily be passed as a separate
argument. Passing thawed data as an argument to a function suffices
to supply an array context. The array context can be extracted from
a thawed argument using, e.g., :func:`~arraycontext.get_container_context`
a thawed argument using, e.g., :func:`~arraycontext.get_container_context_opt`
or :func:`~arraycontext.get_container_context_recursively`.
What does this mean concretely?
......@@ -70,13 +72,75 @@ actual array contexts:
an array expression that has been built up by the user
(using, e.g. :func:`pytato.generate_loopy`).
The interface of an array context
---------------------------------
.. currentmodule:: arraycontext
The :class:`ArrayContext` Interface
-----------------------------------
.. autoclass:: ArrayContext
.. autofunction:: tag_axes
Types and Type Variables for Arrays and Containers
--------------------------------------------------
.. autoclass:: Array
.. autodata:: ArrayT
A type variable with a lower bound of :class:`Array`.
.. autodata:: ScalarLike
A type annotation for scalar types commonly usable with arrays.
See also :class:`ArrayContainer` and :class:`ArrayOrContainerT`.
.. autodata:: ArrayOrContainer
.. autodata:: ArrayOrContainerT
A type variable with a bound of :class:`ArrayOrContainer`.
.. autodata:: ArrayOrArithContainer
.. autodata:: ArrayOrArithContainerT
A type variable with a bound of :class:`ArrayOrArithContainer`.
.. autodata:: ArrayOrArithContainerOrScalar
.. autodata:: ArrayOrArithContainerOrScalarT
A type variable with a bound of :class:`ArrayOrContainerOrScalar`.
.. autodata:: ArrayOrContainerOrScalar
.. autodata:: ArrayOrContainerOrScalarT
A type variable with a bound of :class:`ArrayOrContainerOrScalar`.
.. currentmodule:: arraycontext.context
Canonical locations for type annotations
----------------------------------------
.. class:: ArrayT
:canonical: arraycontext.ArrayT
.. class:: ArrayOrContainerT
:canonical: arraycontext.ArrayOrContainerT
.. class:: ArrayOrContainerOrScalarT
:canonical: arraycontext.ArrayOrContainerOrScalarT
"""
from __future__ import annotations
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
......@@ -102,12 +166,119 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from typing import Sequence, Union, Callable, Any, Tuple
from abc import ABC, abstractmethod, abstractproperty
from abc import ABC, abstractmethod
from collections.abc import Callable, Mapping
from typing import TYPE_CHECKING, Any, Protocol, TypeAlias, TypeVar, Union, overload
from warnings import warn
import numpy as np
from typing_extensions import Self
from pytools import memoize_method
from pytools.tag import Tag
from pytools.tag import ToTagSetConvertible
if TYPE_CHECKING:
import loopy
from arraycontext.container import ArithArrayContainer, ArrayContainer
# {{{ typing
class Array(Protocol):
"""A :class:`~typing.Protocol` for the array type supported by
:class:`ArrayContext`.
This is meant to aid in typing annotations. For a explicit list of
supported types see :attr:`ArrayContext.array_types`.
.. attribute:: shape
.. attribute:: size
.. attribute:: dtype
.. attribute:: __getitem__
In addition, arrays are expected to support basic arithmetic.
"""
@property
def shape(self) -> tuple[int, ...]:
...
@property
def size(self) -> int:
...
@property
def dtype(self) -> np.dtype[Any]:
...
# Covering all the possible index variations is hard and (kind of) futile.
# If you'd like to see how, try changing the Any to
# AxisIndex = slice | int | "Array"
# Index = AxisIndex |tuple[AxisIndex]
def __getitem__(self, index: Any) -> Array:
...
# some basic arithmetic that's supposed to work
def __neg__(self) -> Self: ...
def __abs__(self) -> Self: ...
def __add__(self, other: Self | ScalarLike) -> Self: ...
def __radd__(self, other: Self | ScalarLike) -> Self: ...
def __sub__(self, other: Self | ScalarLike) -> Self: ...
def __rsub__(self, other: Self | ScalarLike) -> Self: ...
def __mul__(self, other: Self | ScalarLike) -> Self: ...
def __rmul__(self, other: Self | ScalarLike) -> Self: ...
def __truediv__(self, other: Self | ScalarLike) -> Self: ...
def __rtruediv__(self, other: Self | ScalarLike) -> Self: ...
# deprecated, use ScalarLike instead
ScalarLike: TypeAlias = int | float | complex | np.generic
Scalar = ScalarLike
ScalarLikeT = TypeVar("ScalarLikeT", bound=ScalarLike)
# NOTE: I'm kind of not sure about the *Tc versions of these type variables.
# mypy seems better at understanding arithmetic performed on the *Tc versions
# than the *T, versions, whereas pyright doesn't seem to care.
#
# This issue seems to be part of it:
# https://github.com/python/mypy/issues/18203
# but there is likely other stuff lurking.
#
# For now, they're purposefully not in the main arraycontext.* name space.
ArrayT = TypeVar("ArrayT", bound=Array)
ArrayOrScalar: TypeAlias = "Array | ScalarLike"
ArrayOrContainer: TypeAlias = "Array | ArrayContainer"
ArrayOrArithContainer: TypeAlias = "Array | ArithArrayContainer"
ArrayOrContainerT = TypeVar("ArrayOrContainerT", bound=ArrayOrContainer)
ArrayOrContainerTc = TypeVar("ArrayOrContainerTc",
Array, "ArrayContainer", "ArithArrayContainer")
ArrayOrArithContainerT = TypeVar("ArrayOrArithContainerT", bound=ArrayOrArithContainer)
ArrayOrArithContainerTc = TypeVar("ArrayOrArithContainerTc",
Array, "ArithArrayContainer")
ArrayOrContainerOrScalar: TypeAlias = "Array | ArrayContainer | ScalarLike"
ArrayOrArithContainerOrScalar: TypeAlias = "Array | ArithArrayContainer | ScalarLike"
ArrayOrContainerOrScalarT = TypeVar(
"ArrayOrContainerOrScalarT",
bound=ArrayOrContainerOrScalar)
ArrayOrArithContainerOrScalarT = TypeVar(
"ArrayOrArithContainerOrScalarT",
bound=ArrayOrArithContainerOrScalar)
ArrayOrContainerOrScalarTc = TypeVar(
"ArrayOrContainerOrScalarTc",
ScalarLike, Array, "ArrayContainer", "ArithArrayContainer")
ArrayOrArithContainerOrScalarTc = TypeVar(
"ArrayOrArithContainerOrScalarTc",
ScalarLike, Array, "ArithArrayContainer")
ContainerOrScalarT = TypeVar("ContainerOrScalarT", bound="ArrayContainer | ScalarLike")
NumpyOrContainerOrScalar = Union[np.ndarray, "ArrayContainer", ScalarLike]
# }}}
# {{{ ArrayContext
......@@ -122,10 +293,6 @@ class ArrayContext(ABC):
.. versionadded:: 2020.2
.. automethod:: empty
.. automethod:: zeros
.. automethod:: empty_like
.. automethod:: zeros_like
.. automethod:: from_numpy
.. automethod:: to_numpy
.. automethod:: call_loopy
......@@ -133,71 +300,102 @@ class ArrayContext(ABC):
.. attribute:: np
Provides access to a namespace that serves as a work-alike to
:mod:`numpy`. The actual level of functionality provided is up to the
:mod:`numpy`. The actual level of functionality provided is up to the
individual array context implementation, however the functions and
objects available under this namespace must not behave differently
from :mod:`numpy`.
As a baseline, special functions available through :mod:`loopy`
(e.g. ``sin``, ``exp``) are accessible through this interface.
A full list of implemented functionality is given in
:ref:`numpy-coverage`.
Callables accessible through this namespace vectorize over object
arrays, including :class:`arraycontext.ArrayContainer`\ s.
.. attribute:: array_types
A :class:`tuple` of types that are the valid base array classes
the context can operate on.
A :class:`tuple` of types that are the valid array classes the
context can operate on. However, it is not necessary that *all* the
:class:`ArrayContext`\ 's operations are legal for the types in
*array_types*. Note that this tuple is *only* intended for use
with :func:`isinstance`. Other uses are not allowed. This allows
for 'types' with overridden :meth:`type.__instancecheck__`.
.. automethod:: freeze
.. automethod:: thaw
.. automethod:: freeze_thaw
.. automethod:: tag
.. automethod:: tag_axis
.. automethod:: compile
"""
array_types: Tuple[type, ...] = ()
array_types: tuple[type, ...] = ()
def __init__(self):
def __init__(self) -> None:
self.np = self._get_fake_numpy_namespace()
def _get_fake_numpy_namespace(self):
from .fake_numpy import BaseFakeNumpyNamespace
return BaseFakeNumpyNamespace(self)
@abstractmethod
def empty(self, shape, dtype):
pass
def _get_fake_numpy_namespace(self) -> Any:
...
@abstractmethod
def zeros(self, shape, dtype):
pass
def __hash__(self) -> int:
raise TypeError(f"unhashable type: '{type(self).__name__}'")
def empty_like(self, ary):
return self.empty(shape=ary.shape, dtype=ary.dtype)
def zeros(self,
shape: int | tuple[int, ...],
dtype: np.dtype[Any]) -> Array:
warn(f"{type(self).__name__}.zeros is deprecated and will stop "
"working in 2025. Use actx.np.zeros instead.",
DeprecationWarning, stacklevel=2)
def zeros_like(self, ary):
return self.zeros(shape=ary.shape, dtype=ary.dtype)
return self.np.zeros(shape, dtype)
@overload
def from_numpy(self, array: np.ndarray) -> Array:
...
@overload
def from_numpy(self, array: ContainerOrScalarT) -> ContainerOrScalarT:
...
@abstractmethod
def from_numpy(self, array: np.ndarray):
def from_numpy(self,
array: NumpyOrContainerOrScalar
) -> ArrayOrContainerOrScalar:
r"""
:returns: the :class:`numpy.ndarray` *array* converted to the
array context's array type. The returned array will be
:meth:`thaw`\ ed.
:meth:`thaw`\ ed. When working with array containers each leaf
must be an :class:`~numpy.ndarray` or scalar, which is then converted
to the context's array type leaving the container structure
intact.
"""
pass
@overload
def to_numpy(self, array: Array) -> np.ndarray:
...
@overload
def to_numpy(self, array: ContainerOrScalarT) -> ContainerOrScalarT:
...
@abstractmethod
def to_numpy(self, array):
def to_numpy(self,
array: ArrayOrContainerOrScalar
) -> NumpyOrContainerOrScalar:
r"""
:returns: *array*, an array recognized by the context, converted
to a :class:`numpy.ndarray`. *array* must be
:meth:`thaw`\ ed.
:returns: an :class:`numpy.ndarray` for each array recognized by the
context. The input *array* must be :meth:`thaw`\ ed.
When working with array containers each leaf must be one of
the context's array types or a scalar, which is then converted to
an :class:`~numpy.ndarray` leaving the container structure intact.
"""
pass
def call_loopy(self, program, **kwargs):
@abstractmethod
def call_loopy(self,
t_unit: loopy.TranslationUnit,
**kwargs: Any) -> dict[str, Array]:
"""Execute the :mod:`loopy` program *program* on the arguments
*kwargs*.
......@@ -210,7 +408,7 @@ class ArrayContext(ABC):
"""
@abstractmethod
def freeze(self, array):
def freeze(self, array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
"""Return a version of the context-defined array *array* that is
'frozen', i.e. suitable for long-term storage and reuse. Frozen arrays
do not support arithmetic. For example, in the context of
......@@ -221,12 +419,10 @@ class ArrayContext(ABC):
Freezing makes the array independent of this :class:`ArrayContext`;
it is permitted to :meth:`thaw` it in a different one, as long as that
context understands the array format.
See also :func:`arraycontext.freeze`.
"""
@abstractmethod
def thaw(self, array):
def thaw(self, array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
"""Take a 'frozen' array and return a new array representing the data in
*array* that is able to perform arithmetic and other operations, using
the execution resources of this context. In the context of
......@@ -236,39 +432,66 @@ class ArrayContext(ABC):
the data in *array*.
The returned array may not be used with other contexts while thawed.
"""
See also :func:`arraycontext.thaw`.
def freeze_thaw(
self, array: ArrayOrContainerOrScalarT
) -> ArrayOrContainerOrScalarT:
r"""Evaluate an input array or container to "frozen" data return a new
"thawed" array or container representing the evaluation result that is
ready for use. This is a shortcut for calling :meth:`freeze` and
:meth:`thaw`.
This method can be useful in array contexts backed by, e.g.
:mod:`pytato`, to force the evaluation of a built-up array expression
(and thereby avoid reevaluations for expressions built on the array).
"""
return self.thaw(self.freeze(array))
@abstractmethod
def tag(self, tags: Union[Sequence[Tag], Tag], array):
def tag(self,
tags: ToTagSetConvertible,
array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
"""If the array type used by the array context is capable of capturing
metadata, return a version of *array* with the *tags* applied. *array*
itself is not modified.
itself is not modified. When working with array containers, the
tags are applied to each leaf of the container.
See :ref:`metadata` as well as application-specific metadata types.
.. versionadded:: 2021.2
"""
@abstractmethod
def tag_axis(self, iaxis, tags: Union[Sequence[Tag], Tag], array):
def tag_axis(self,
iaxis: int, tags: ToTagSetConvertible,
array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
"""If the array type used by the array context is capable of capturing
metadata, return a version of *array* in which axis number *iaxis* has
the *tags* applied. *array* itself is not modified.
the *tags* applied. *array* itself is not modified. When working with
array containers, the tags are applied to each leaf of the container.
See :ref:`metadata` as well as application-specific metadata types.
.. versionadded:: 2021.2
"""
@memoize_method
def _get_einsum_prg(self, spec, arg_names, tagged):
def _get_einsum_prg(self,
spec: str, arg_names: tuple[str, ...],
tagged: ToTagSetConvertible) -> loopy.TranslationUnit:
import loopy as lp
from .loopy import _DEFAULT_LOOPY_OPTIONS
from loopy.version import MOST_RECENT_LANGUAGE_VERSION
from .loopy import _DEFAULT_LOOPY_OPTIONS
return lp.make_einsum(
spec,
arg_names,
options=_DEFAULT_LOOPY_OPTIONS,
lang_version=MOST_RECENT_LANGUAGE_VERSION,
tags=tagged,
default_order=lp.auto,
default_offset=lp.auto,
)
# This lives here rather than in .np because the interface does not
......@@ -284,7 +507,10 @@ class ArrayContext(ABC):
# That's why einsum's interface here needs to be cluttered with
# metadata, and that's why it can't live under .np.
# [1] https://github.com/inducer/meshmode/issues/177
def einsum(self, spec, *args, arg_names=None, tagged=()):
def einsum(self,
spec: str, *args: Array,
arg_names: tuple[str, ...] | None = None,
tagged: ToTagSetConvertible = ()) -> Array:
"""Computes the result of Einstein summation following the
convention in :func:`numpy.einsum`.
......@@ -304,15 +530,16 @@ class ArrayContext(ABC):
:return: the output of the einsum :mod:`loopy` program
"""
if arg_names is None:
arg_names = tuple("arg%d" % i for i in range(len(args)))
arg_names = tuple(f"arg{i}" for i in range(len(args)))
prg = self._get_einsum_prg(spec, arg_names, tagged)
return self.call_loopy(
out_ary = self.call_loopy(
prg, **{arg_names[i]: arg for i, arg in enumerate(args)}
)["out"]
return self.tag(tagged, out_ary)
@abstractmethod
def clone(self):
def clone(self) -> Self:
"""If possible, return a version of *self* that is semantically
equivalent (i.e. implements all array operations in the same way)
but is a separate object. May return *self* if that is not possible.
......@@ -351,10 +578,52 @@ class ArrayContext(ABC):
return f
# undocumented for now
@abstractproperty
def permits_inplace_modification(self):
pass
@property
@abstractmethod
def permits_inplace_modification(self) -> bool:
"""
*True* if the arrays allow in-place modifications.
"""
# undocumented for now
@property
@abstractmethod
def supports_nonscalar_broadcasting(self) -> bool:
"""
*True* if the arrays support non-scalar broadcasting.
"""
# undocumented for now
@property
@abstractmethod
def permits_advanced_indexing(self) -> bool:
"""
*True* if the arrays support :mod:`numpy`'s advanced indexing semantics.
"""
# }}}
# {{{ tagging helpers
def tag_axes(
actx: ArrayContext,
dim_to_tags: Mapping[int, ToTagSetConvertible],
ary: ArrayT) -> ArrayT:
"""
Return a copy of *ary* with the axes in *dim_to_tags* tagged with their
corresponding tags. Equivalent to repeated application of
:meth:`ArrayContext.tag_axis`.
"""
for iaxis, tags in dim_to_tags.items():
ary = actx.tag_axis(iaxis, tags, ary)
return ary
# }}}
class UntransformedCodeWarning(UserWarning):
pass
# vim: foldmethod=marker
arraycontext/fake_numpy.py
View file @ 21a89f1b
from __future__ import annotations
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
"""
......@@ -23,59 +26,25 @@ THE SOFTWARE.
"""
import operator
from abc import ABC, abstractmethod
from typing import Any
import numpy as np
from arraycontext.container import is_array_container, serialize_container
from arraycontext.container.traversal import (
rec_map_array_container, multimapped_over_array_containers)
from pytools import memoize_in
# {{{ _get_scalar_func_loopy_program
def _get_scalar_func_loopy_program(actx, c_name, nargs, naxes):
@memoize_in(actx, _get_scalar_func_loopy_program)
def get(c_name, nargs, naxes):
from pymbolic import var
var_names = ["i%d" % i for i in range(naxes)]
size_names = ["n%d" % i for i in range(naxes)]
subscript = tuple(var(vname) for vname in var_names)
from islpy import make_zero_and_vars
v = make_zero_and_vars(var_names, params=size_names)
domain = v[0].domain()
for vname, sname in zip(var_names, size_names):
domain = domain & v[0].le_set(v[vname]) & v[vname].lt_set(v[sname])
domain_bset, = domain.get_basic_sets()
import loopy as lp
from .loopy import make_loopy_program
from arraycontext.transform_metadata import ElementwiseMapKernelTag
return make_loopy_program(
[domain_bset],
[
lp.Assignment(
var("out")[subscript],
var(c_name)(*[
var("inp%d" % i)[subscript] for i in range(nargs)]))
],
name="actx_special_%s" % c_name,
tags=(ElementwiseMapKernelTag(),))
return get(c_name, nargs, naxes)
# }}}
from arraycontext.container import NotAnArrayContainerError, serialize_container
from arraycontext.container.traversal import rec_map_array_container
# {{{ BaseFakeNumpyNamespace
class BaseFakeNumpyNamespace:
class BaseFakeNumpyNamespace(ABC):
def __init__(self, array_context):
self._array_context = array_context
self.linalg = self._get_fake_numpy_linalg_namespace()
def _get_fake_numpy_linalg_namespace(self):
return BaseFakeNumpyLinalgNamespace(self.array_context)
return BaseFakeNumpyLinalgNamespace(self._array_context)
_numpy_math_functions = frozenset({
# https://numpy.org/doc/stable/reference/routines.math.html
......@@ -124,74 +93,20 @@ class BaseFakeNumpyNamespace:
# Miscellaneous
"convolve", "clip", "sqrt", "cbrt", "square", "absolute", "abs", "fabs",
"sign", "heaviside", "maximum", "fmax", "nan_to_num",
"sign", "heaviside", "maximum", "fmax", "nan_to_num", "isnan", "minimum",
"fmin",
# FIXME:
# "interp",
})
_numpy_to_c_arc_functions = {
"arcsin": "asin",
"arccos": "acos",
"arctan": "atan",
"arctan2": "atan2",
"arcsinh": "asinh",
"arccosh": "acosh",
"arctanh": "atanh",
}
_c_to_numpy_arc_functions = {c_name: numpy_name
for numpy_name, c_name in _numpy_to_c_arc_functions.items()}
def __getattr__(self, name):
def loopy_implemented_elwise_func(*args):
actx = self._array_context
prg = _get_scalar_func_loopy_program(actx,
c_name, nargs=len(args), naxes=len(args[0].shape))
outputs = actx.call_loopy(prg,
**{"inp%d" % i: arg for i, arg in enumerate(args)})
return outputs["out"]
if name in self._c_to_numpy_arc_functions:
from warnings import warn
warn(f"'{name}' in ArrayContext.np is deprecated. "
"Use '{c_to_numpy_arc_functions[name]}' as in numpy. "
"The old name will stop working in 2021.",
DeprecationWarning, stacklevel=3)
# normalize to C names anyway
c_name = self._numpy_to_c_arc_functions.get(name, name)
# limit which functions we try to hand off to loopy
if name in self._numpy_math_functions:
return multimapped_over_array_containers(loopy_implemented_elwise_func)
else:
raise AttributeError(name)
def _new_like(self, ary, alloc_like):
from numbers import Number
if isinstance(ary, np.ndarray) and ary.dtype.char == "O":
# NOTE: we don't want to match numpy semantics on object arrays,
# e.g. `np.zeros_like(x)` returns `array([0, 0, ...], dtype=object)`
# FIXME: what about object arrays nested in an ArrayContainer?
raise NotImplementedError("operation not implemented for object arrays")
elif is_array_container(ary):
return rec_map_array_container(alloc_like, ary)
elif isinstance(ary, Number):
# NOTE: `np.zeros_like(x)` returns `array(x, shape=())`, which
# is best implemented by concrete array contexts, if at all
raise NotImplementedError("operation not implemented for scalars")
else:
return alloc_like(ary)
def empty_like(self, ary):
return self._new_like(ary, self._array_context.empty_like)
@abstractmethod
def zeros(self, shape, dtype):
...
@abstractmethod
def zeros_like(self, ary):
return self._new_like(ary, self._array_context.zeros_like)
...
def conjugate(self, x):
# NOTE: conjugate distributes over object arrays, but it looks for a
......@@ -201,22 +116,110 @@ class BaseFakeNumpyNamespace:
conj = conjugate
# {{{ linspace
# based on
# https://github.com/numpy/numpy/blob/v1.25.0/numpy/core/function_base.py#L24-L182
def linspace(self, start, stop, num=50, endpoint=True, retstep=False, dtype=None,
axis=0):
num = operator.index(num)
if num < 0:
raise ValueError(f"Number of samples, {num}, must be non-negative.")
div = (num - 1) if endpoint else num
# Convert float/complex array scalars to float, gh-3504
# and make sure one can use variables that have an __array_interface__,
# gh-6634
if isinstance(start, self._array_context.array_types):
raise NotImplementedError("start as an actx array")
if isinstance(stop, self._array_context.array_types):
raise NotImplementedError("stop as an actx array")
start = np.array(start) * 1.0
stop = np.array(stop) * 1.0
dt = np.result_type(start, stop, float(num))
if dtype is None:
dtype = dt
integer_dtype = False
else:
integer_dtype = np.issubdtype(dtype, np.integer)
delta = stop - start
y = self.arange(0, num, dtype=dt).reshape((-1,) + (1,) * delta.ndim)
if div > 0:
step = delta / div
# any_step_zero = _nx.asanyarray(step == 0).any()
any_step_zero = self._array_context.to_numpy(step == 0).any()
if any_step_zero:
delta_actx = self._array_context.from_numpy(delta)
# Special handling for denormal numbers, gh-5437
y = y / div
y = y * delta_actx
else:
step_actx = self._array_context.from_numpy(step)
y = y * step_actx
else:
delta_actx = self._array_context.from_numpy(delta)
# sequences with 0 items or 1 item with endpoint=True (i.e. div <= 0)
# have an undefined step
step = np.nan
# Multiply with delta to allow possible override of output class.
y = y * delta_actx
y += start
# FIXME reenable, without in-place ops
# if endpoint and num > 1:
# y[-1, ...] = stop
if axis != 0:
# y = _nx.moveaxis(y, 0, axis)
raise NotImplementedError("axis != 0")
if integer_dtype:
y = self.floor(y) # pylint: disable=no-member
# FIXME: Use astype
# https://github.com/inducer/pytato/issues/456
if retstep:
return y, step
# return y.astype(dtype), step
else:
return y
# return y.astype(dtype)
# }}}
def arange(self, *args: Any, **kwargs: Any):
raise NotImplementedError
# }}}
# {{{ BaseFakeNumpyLinalgNamespace
def _scalar_list_norm(ary, ord):
def _reduce_norm(actx, arys, ord):
from functools import reduce
from numbers import Number
if ord is None:
ord = 2
from numbers import Number
if ord == np.inf:
return max(ary)
# NOTE: these are ordered by an expected usage frequency
if ord == 2:
return actx.np.sqrt(sum(subary*subary for subary in arys))
elif ord == np.inf:
return reduce(actx.np.maximum, arys)
elif ord == -np.inf:
return min(ary)
return reduce(actx.np.minimum, arys)
elif isinstance(ord, Number) and ord > 0:
return sum(iary**ord for iary in ary)**(1/ord)
return sum(subary**ord for subary in arys)**(1/ord)
else:
raise NotImplementedError(f"unsupported value of 'ord': {ord}")
......@@ -226,9 +229,7 @@ class BaseFakeNumpyLinalgNamespace:
self._array_context = array_context
def norm(self, ary, ord=None):
from numbers import Number
if isinstance(ary, Number):
if np.isscalar(ary):
return abs(ary)
actx = self._array_context
......@@ -249,10 +250,13 @@ class BaseFakeNumpyLinalgNamespace:
return flat_norm(ary, ord=ord)
if is_array_container(ary):
return _scalar_list_norm([
self.norm(subary, ord=ord)
for _, subary in serialize_container(ary)
try:
iterable = serialize_container(ary)
except NotAnArrayContainerError:
pass
else:
return _reduce_norm(actx, [
self.norm(subary, ord=ord) for _, subary in iterable
], ord=ord)
if ord is None:
......@@ -262,16 +266,20 @@ class BaseFakeNumpyLinalgNamespace:
raise NotImplementedError("only vector norms are implemented")
if ary.size == 0:
return 0
return ary.dtype.type(0)
from numbers import Number
if ord == 2:
return actx.np.sqrt(actx.np.sum(abs(ary)**2))
if ord == np.inf:
return self._array_context.np.max(abs(ary))
return actx.np.max(abs(ary))
elif ord == -np.inf:
return self._array_context.np.min(abs(ary))
return actx.np.min(abs(ary))
elif isinstance(ord, Number) and ord > 0:
return self._array_context.np.sum(abs(ary)**ord)**(1/ord)
return actx.np.sum(abs(ary)**ord)**(1/ord)
else:
raise NotImplementedError(f"unsupported value of 'ord': {ord}")
# }}}
......
arraycontext/impl/__init__.py
View file @ 21a89f1b
from __future__ import annotations
__copyright__ = """
Copyright (C) 2020-1 University of Illinois Board of Trustees
"""
......
arraycontext/impl/jax/__init__.py 0 → 100644
View file @ 21a89f1b
"""
.. currentmodule:: arraycontext
.. autoclass:: EagerJAXArrayContext
"""
from __future__ import annotations
__copyright__ = """
Copyright (C) 2021 University of Illinois Board of Trustees
"""
__license__ = """
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.
"""
from collections.abc import Callable
import numpy as np
from pytools.tag import ToTagSetConvertible
from arraycontext.container.traversal import rec_map_array_container, with_array_context
from arraycontext.context import Array, ArrayContext, ArrayOrContainer, ScalarLike
class EagerJAXArrayContext(ArrayContext):
"""
A :class:`ArrayContext` that uses
:class:`jax.Array` instances for its base array
class and performs all array operations eagerly. See
:class:`~arraycontext.PytatoJAXArrayContext` for a lazier version.
.. note::
JAX stores a global configuration state in :data:`jax.config`. Callers
are expected to maintain those. Most important for scientific computing
workloads being ``jax_enable_x64``.
"""
def __init__(self) -> None:
super().__init__()
import jax.numpy as jnp
self.array_types = (jnp.ndarray, )
def _get_fake_numpy_namespace(self):
from .fake_numpy import EagerJAXFakeNumpyNamespace
return EagerJAXFakeNumpyNamespace(self)
def _rec_map_container(
self, func: Callable[[Array], Array], array: ArrayOrContainer,
allowed_types: tuple[type, ...] | None = None, *,
default_scalar: ScalarLike | None = None,
strict: bool = False) -> ArrayOrContainer:
if allowed_types is None:
allowed_types = self.array_types
def _wrapper(ary):
if isinstance(ary, allowed_types):
return func(ary)
elif np.isscalar(ary):
if default_scalar is None:
return ary
else:
return np.array(ary).dtype.type(default_scalar)
else:
raise TypeError(
f"{type(self).__name__}.{func.__name__[1:]} invoked with "
f"an unsupported array type: got '{type(ary).__name__}', "
f"but expected one of {allowed_types}")
return rec_map_array_container(_wrapper, array)
# {{{ ArrayContext interface
def from_numpy(self, array):
def _from_numpy(ary):
import jax
return jax.device_put(ary)
return with_array_context(
self._rec_map_container(_from_numpy, array, allowed_types=(np.ndarray,)),
actx=self)
def to_numpy(self, array):
def _to_numpy(ary):
import jax
return jax.device_get(ary)
return with_array_context(
self._rec_map_container(_to_numpy, array),
actx=None)
def freeze(self, array):
def _freeze(ary):
return ary.block_until_ready()
return with_array_context(self._rec_map_container(_freeze, array), actx=None)
def thaw(self, array):
return with_array_context(array, actx=self)
def tag(self, tags: ToTagSetConvertible, array):
# Sorry, not capable.
return array
def tag_axis(self, iaxis, tags: ToTagSetConvertible, array):
# TODO: See `jax.experimental.maps.xmap`, probably that should be useful?
return array
def call_loopy(self, t_unit, **kwargs):
raise NotImplementedError(
"Calling loopy on JAX arrays is not supported. Maybe rewrite"
" the loopy kernel as numpy-flavored array operations using"
" ArrayContext.np.")
def einsum(self, spec, *args, arg_names=None, tagged=()):
import jax.numpy as jnp
if arg_names is not None:
from warnings import warn
warn("'arg_names' don't bear any significance in "
f"{type(self).__name__}.", stacklevel=2)
return jnp.einsum(spec, *args)
def clone(self):
return type(self)()
# }}}
# {{{ properties
@property
def permits_inplace_modification(self):
return False
@property
def supports_nonscalar_broadcasting(self):
return True
@property
def permits_advanced_indexing(self):
return True
# }}}
# vim: foldmethod=marker
arraycontext/impl/jax/fake_numpy.py 0 → 100644
View file @ 21a89f1b
from __future__ import annotations
__copyright__ = """
Copyright (C) 2021 University of Illinois Board of Trustees
"""
__license__ = """
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.
"""
from functools import partial, reduce
import numpy as np
import jax.numpy as jnp
from arraycontext.container import (
NotAnArrayContainerError,
serialize_container,
)
from arraycontext.container.traversal import (
rec_map_array_container,
rec_map_reduce_array_container,
rec_multimap_array_container,
)
from arraycontext.context import Array, ArrayOrContainer
from arraycontext.fake_numpy import BaseFakeNumpyLinalgNamespace, BaseFakeNumpyNamespace
class EagerJAXFakeNumpyLinalgNamespace(BaseFakeNumpyLinalgNamespace):
# Everything is implemented in the base class for now.
pass
class EagerJAXFakeNumpyNamespace(BaseFakeNumpyNamespace):
"""
A :mod:`numpy` mimic for :class:`~arraycontext.EagerJAXArrayContext`.
"""
def _get_fake_numpy_linalg_namespace(self):
return EagerJAXFakeNumpyLinalgNamespace(self._array_context)
def __getattr__(self, name):
return partial(rec_multimap_array_container, getattr(jnp, name))
# NOTE: the order of these follows the order in numpy docs
# NOTE: when adding a function here, also add it to `array_context.rst` docs!
# {{{ array creation routines
def zeros(self, shape, dtype):
return jnp.zeros(shape=shape, dtype=dtype)
def empty_like(self, ary):
from warnings import warn
warn(f"{type(self._array_context).__name__}.np.empty_like is "
"deprecated and will stop working in 2023. Prefer actx.np.zeros_like "
"instead.",
DeprecationWarning, stacklevel=2)
def _empty_like(array):
return self._array_context.empty(array.shape, array.dtype)
return self._array_context._rec_map_container(_empty_like, ary)
def zeros_like(self, ary):
def _zeros_like(array):
return self._array_context.zeros(array.shape, array.dtype)
return self._array_context._rec_map_container(
_zeros_like, ary, default_scalar=0)
def ones_like(self, ary):
return self.full_like(ary, 1)
def full_like(self, ary, fill_value):
def _full_like(subary):
return jnp.full_like(subary, fill_value)
return self._array_context._rec_map_container(
_full_like, ary, default_scalar=fill_value)
# }}}
# {{{ array manipulation routies
def reshape(self, a, newshape, order="C"):
return rec_map_array_container(
lambda ary: jnp.reshape(ary, newshape, order=order),
a)
def ravel(self, a, order="C"):
"""
.. warning::
Since :func:`jax.numpy.reshape` does not support orders `A`` and
``K``, in such cases we fallback to using ``order = C``.
"""
if order in "AK":
from warnings import warn
warn(f"ravel with order='{order}' not supported by JAX,"
" using order=C.", stacklevel=1)
order = "C"
return rec_map_array_container(
lambda subary: jnp.ravel(subary, order=order), a)
def transpose(self, a, axes=None):
return rec_multimap_array_container(jnp.transpose, a, axes)
def broadcast_to(self, array, shape):
return rec_map_array_container(partial(jnp.broadcast_to, shape=shape), array)
def concatenate(self, arrays, axis=0):
return rec_multimap_array_container(jnp.concatenate, arrays, axis)
def stack(self, arrays, axis=0):
return rec_multimap_array_container(
lambda *args: jnp.stack(arrays=args, axis=axis),
*arrays)
# }}}
# {{{ linear algebra
def vdot(self, x, y, dtype=None):
from arraycontext import rec_multimap_reduce_array_container
def _rec_vdot(ary1, ary2):
common_dtype = np.result_type(ary1, ary2)
if dtype not in (None, common_dtype):
raise NotImplementedError(
f"{type(self).__name__} cannot take dtype in vdot.")
return jnp.vdot(ary1, ary2)
return rec_multimap_reduce_array_container(sum, _rec_vdot, x, y)
# }}}
# {{{ logic functions
def all(self, a):
return rec_map_reduce_array_container(
partial(reduce, jnp.logical_and), jnp.all, a)
def any(self, a):
return rec_map_reduce_array_container(
partial(reduce, jnp.logical_or), jnp.any, a)
def array_equal(self, a: ArrayOrContainer, b: ArrayOrContainer) -> Array:
actx = self._array_context
# NOTE: not all backends support `bool` properly, so use `int8` instead
true_ary = actx.from_numpy(np.int8(True))
false_ary = actx.from_numpy(np.int8(False))
def rec_equal(x, y):
if type(x) is not type(y):
return false_ary
try:
serialized_x = serialize_container(x)
serialized_y = serialize_container(y)
except NotAnArrayContainerError:
if x.shape != y.shape:
return false_ary
else:
return jnp.all(jnp.equal(x, y))
else:
if len(serialized_x) != len(serialized_y):
return false_ary
return reduce(
jnp.logical_and,
[(true_ary if kx_i == ky_i else false_ary)
and rec_equal(x_i, y_i)
for (kx_i, x_i), (ky_i, y_i)
in zip(serialized_x, serialized_y, strict=True)],
true_ary)
return rec_equal(a, b)
# }}}
# {{{ mathematical functions
def sum(self, a, axis=None, dtype=None):
return rec_map_reduce_array_container(
sum,
partial(jnp.sum, axis=axis, dtype=dtype),
a)
def amin(self, a, axis=None):
return rec_map_reduce_array_container(
partial(reduce, jnp.minimum), partial(jnp.amin, axis=axis), a)
min = amin
def amax(self, a, axis=None):
return rec_map_reduce_array_container(
partial(reduce, jnp.maximum), partial(jnp.amax, axis=axis), a)
max = amax
# }}}
# {{{ sorting, searching and counting
def where(self, criterion, then, else_):
return rec_multimap_array_container(jnp.where, criterion, then, else_)
# }}}
arraycontext/impl/numpy/__init__.py 0 → 100644
View file @ 21a89f1b
"""
.. currentmodule:: arraycontext
A :mod:`numpy`-based array context.
.. autoclass:: NumpyArrayContext
"""
from __future__ import annotations
__copyright__ = """
Copyright (C) 2021 University of Illinois Board of Trustees
"""
__license__ = """
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.
"""
from typing import Any, overload
import numpy as np
import loopy as lp
from pytools.tag import ToTagSetConvertible
from arraycontext.container.traversal import rec_map_array_container, with_array_context
from arraycontext.context import (
Array,
ArrayContext,
ArrayOrContainerOrScalar,
ArrayOrContainerOrScalarT,
ContainerOrScalarT,
NumpyOrContainerOrScalar,
UntransformedCodeWarning,
)
class NumpyNonObjectArrayMetaclass(type):
def __instancecheck__(cls, instance: Any) -> bool:
return isinstance(instance, np.ndarray) and instance.dtype != object
class NumpyNonObjectArray(metaclass=NumpyNonObjectArrayMetaclass):
pass
class NumpyArrayContext(ArrayContext):
"""
A :class:`ArrayContext` that uses :class:`numpy.ndarray` to represent arrays.
.. automethod:: __init__
"""
_loopy_transform_cache: dict[lp.TranslationUnit, lp.ExecutorBase]
def __init__(self) -> None:
super().__init__()
self._loopy_transform_cache = {}
array_types = (NumpyNonObjectArray,)
def _get_fake_numpy_namespace(self):
from .fake_numpy import NumpyFakeNumpyNamespace
return NumpyFakeNumpyNamespace(self)
# {{{ ArrayContext interface
def clone(self):
return type(self)()
@overload
def from_numpy(self, array: np.ndarray) -> Array:
...
@overload
def from_numpy(self, array: ContainerOrScalarT) -> ContainerOrScalarT:
...
def from_numpy(self,
array: NumpyOrContainerOrScalar
) -> ArrayOrContainerOrScalar:
return array
@overload
def to_numpy(self, array: Array) -> np.ndarray:
...
@overload
def to_numpy(self, array: ContainerOrScalarT) -> ContainerOrScalarT:
...
def to_numpy(self,
array: ArrayOrContainerOrScalar
) -> NumpyOrContainerOrScalar:
return array
def call_loopy(
self,
t_unit: lp.TranslationUnit, **kwargs: Any
) -> dict[str, Array]:
t_unit = t_unit.copy(target=lp.ExecutableCTarget())
try:
executor = self._loopy_transform_cache[t_unit]
except KeyError:
executor = self.transform_loopy_program(t_unit).executor()
self._loopy_transform_cache[t_unit] = executor
_, result = executor(**kwargs)
return result
def freeze(self, array):
def _freeze(ary):
return ary
return with_array_context(rec_map_array_container(_freeze, array), actx=None)
def thaw(self, array):
def _thaw(ary):
return ary
return with_array_context(rec_map_array_container(_thaw, array), actx=self)
# }}}
def transform_loopy_program(self, t_unit):
from warnings import warn
warn("Using the base "
f"{type(self).__name__}.transform_loopy_program "
"to transform a translation unit. "
"This is a no-op and will result in unoptimized C code for"
"the requested optimization, all in a single statement."
"This will work, but is unlikely to be performant."
f"Instead, subclass {type(self).__name__} and implement "
"the specific transform logic required to transform the program "
"for your package or application. Check higher-level packages "
"(e.g. meshmode), which may already have subclasses you may want "
"to build on.",
UntransformedCodeWarning, stacklevel=2)
return t_unit
def tag(self,
tags: ToTagSetConvertible,
array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
# Numpy doesn't support tagging
return array
def tag_axis(self,
iaxis: int, tags: ToTagSetConvertible,
array: ArrayOrContainerOrScalarT) -> ArrayOrContainerOrScalarT:
# Numpy doesn't support tagging
return array
def einsum(self, spec, *args, arg_names=None, tagged=()):
return np.einsum(spec, *args)
@property
def permits_inplace_modification(self):
return True
@property
def supports_nonscalar_broadcasting(self):
return True
@property
def permits_advanced_indexing(self):
return True
arraycontext/impl/numpy/fake_numpy.py 0 → 100644
View file @ 21a89f1b
from __future__ import annotations
__copyright__ = """
Copyright (C) 2021 University of Illinois Board of Trustees
"""
__license__ = """
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.
"""
from functools import partial, reduce
from typing import cast
import numpy as np
from arraycontext.container import NotAnArrayContainerError, serialize_container
from arraycontext.container.traversal import (
rec_map_array_container,
rec_map_reduce_array_container,
rec_multimap_array_container,
rec_multimap_reduce_array_container,
)
from arraycontext.context import Array, ArrayOrContainer
from arraycontext.fake_numpy import (
BaseFakeNumpyLinalgNamespace,
BaseFakeNumpyNamespace,
)
class NumpyFakeNumpyLinalgNamespace(BaseFakeNumpyLinalgNamespace):
# Everything is implemented in the base class for now.
pass
_NUMPY_UFUNCS = frozenset({"concatenate", "reshape", "transpose",
"ones_like", "where",
*BaseFakeNumpyNamespace._numpy_math_functions
})
class NumpyFakeNumpyNamespace(BaseFakeNumpyNamespace):
"""
A :mod:`numpy` mimic for :class:`NumpyArrayContext`.
"""
def _get_fake_numpy_linalg_namespace(self):
return NumpyFakeNumpyLinalgNamespace(self._array_context)
def zeros(self, shape, dtype):
return np.zeros(shape, dtype)
def __getattr__(self, name):
if name in _NUMPY_UFUNCS:
from functools import partial
return partial(rec_multimap_array_container,
getattr(np, name))
raise AttributeError(name)
def sum(self, a, axis=None, dtype=None):
return rec_map_reduce_array_container(sum, partial(np.sum,
axis=axis,
dtype=dtype),
a)
def min(self, a, axis=None):
return rec_map_reduce_array_container(
partial(reduce, np.minimum), partial(np.amin, axis=axis), a)
def max(self, a, axis=None):
return rec_map_reduce_array_container(
partial(reduce, np.maximum), partial(np.amax, axis=axis), a)
def stack(self, arrays, axis=0):
return rec_multimap_array_container(
lambda *args: np.stack(arrays=args, axis=axis),
*arrays)
def broadcast_to(self, array, shape):
return rec_map_array_container(partial(np.broadcast_to, shape=shape), array)
# {{{ relational operators
def equal(self, x, y):
return rec_multimap_array_container(np.equal, x, y)
def not_equal(self, x, y):
return rec_multimap_array_container(np.not_equal, x, y)
def greater(self, x, y):
return rec_multimap_array_container(np.greater, x, y)
def greater_equal(self, x, y):
return rec_multimap_array_container(np.greater_equal, x, y)
def less(self, x, y):
return rec_multimap_array_container(np.less, x, y)
def less_equal(self, x, y):
return rec_multimap_array_container(np.less_equal, x, y)
# }}}
def ravel(self, a, order="C"):
return rec_map_array_container(partial(np.ravel, order=order), a)
def vdot(self, x, y):
return rec_multimap_reduce_array_container(sum, np.vdot, x, y)
def any(self, a):
return rec_map_reduce_array_container(partial(reduce, np.logical_or),
lambda subary: np.any(subary), a)
def all(self, a):
return rec_map_reduce_array_container(partial(reduce, np.logical_and),
lambda subary: np.all(subary), a)
def array_equal(self, a: ArrayOrContainer, b: ArrayOrContainer) -> Array:
false_ary = np.array(False)
true_ary = np.array(True)
if type(a) is not type(b):
return false_ary
try:
serialized_x = serialize_container(a)
serialized_y = serialize_container(b)
except NotAnArrayContainerError:
assert isinstance(a, np.ndarray)
assert isinstance(b, np.ndarray)
return np.array(np.array_equal(a, b))
else:
if len(serialized_x) != len(serialized_y):
return false_ary
return np.logical_and.reduce(
[(true_ary if kx_i == ky_i else false_ary)
and cast(np.ndarray, self.array_equal(x_i, y_i))
for (kx_i, x_i), (ky_i, y_i)
in zip(serialized_x, serialized_y, strict=True)],
initial=true_ary)
def arange(self, *args, **kwargs):
return np.arange(*args, **kwargs)
def linspace(self, *args, **kwargs):
return np.linspace(*args, **kwargs)
def zeros_like(self, ary):
return rec_map_array_container(np.zeros_like, ary)
def reshape(self, a, newshape, order="C"):
return rec_map_array_container(
lambda ary: ary.reshape(newshape, order=order),
a)
# vim: fdm=marker
arraycontext/impl/pyopencl/__init__.py
View file @ 21a89f1b
"""
from __future__ import annotations
__doc__ = """
.. currentmodule:: arraycontext
.. autoclass:: PyOpenCLArrayContext
.. automodule:: arraycontext.impl.pyopencl.taggable_cl_array
"""
__copyright__ = """
......@@ -27,19 +31,27 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""
from collections.abc import Callable
from typing import TYPE_CHECKING
from warnings import warn
from typing import Dict, List, Sequence, Optional, Union, TYPE_CHECKING
import numpy as np
from pytools.tag import Tag
from pytools.tag import ToTagSetConvertible
from arraycontext.context import ArrayContext
from arraycontext.container.traversal import rec_map_array_container, with_array_context
from arraycontext.context import (
Array,
ArrayContext,
ArrayOrContainer,
ScalarLike,
UntransformedCodeWarning,
)
if TYPE_CHECKING:
import pyopencl
import loopy as lp
import pyopencl
# {{{ PyOpenCLArrayContext
......@@ -65,13 +77,15 @@ class PyOpenCLArrayContext(ArrayContext):
of arrays are created (e.g. as results of computation), the associated cost
may become significant. Using e.g. :class:`pyopencl.tools.MemoryPool`
as the allocator can help avoid this cost.
.. automethod:: transform_loopy_program
"""
def __init__(self,
queue: "pyopencl.CommandQueue",
allocator: Optional["pyopencl.tools.AllocatorInterface"] = None,
wait_event_queue_length: Optional[int] = None,
force_device_scalars: bool = False) -> None:
queue: pyopencl.CommandQueue,
allocator: pyopencl.tools.AllocatorBase | None = None,
wait_event_queue_length: int | None = None,
force_device_scalars: bool | None = None) -> None:
r"""
:arg wait_event_queue_length: The length of a queue of
:class:`~pyopencl.Event` objects that are maintained by the
......@@ -92,24 +106,18 @@ class PyOpenCLArrayContext(ArrayContext):
For now, *wait_event_queue_length* should be regarded as an
experimental feature that may change or disappear at any minute.
:arg force_device_scalars: if *True*, scalar results returned from
reductions in :attr:`ArrayContext.np` will be kept on the device.
If *False*, the equivalent of :meth:`~ArrayContext.freeze` and
:meth:`~ArrayContext.to_numpy` is applied to transfer the results
to the host.
"""
if not force_device_scalars:
warn("Configuring the PyOpenCLArrayContext to return host scalars "
"from reductions is deprecated. "
"To configure the PyOpenCLArrayContext to return "
"device scalars, pass 'force_device_scalars=True' to the "
"constructor. "
"Support for returning host scalars will be removed in 2022.",
DeprecationWarning, stacklevel=2)
if force_device_scalars is not None:
warn(
"`force_device_scalars` is deprecated and will be removed in 2025.",
DeprecationWarning, stacklevel=2)
if not force_device_scalars:
raise ValueError(
"Passing force_device_scalars=False is not allowed.")
import pyopencl as cl
import pyopencl.array as cla
import pyopencl.array as cl_array
super().__init__()
self.context = queue.context
......@@ -118,67 +126,134 @@ class PyOpenCLArrayContext(ArrayContext):
if wait_event_queue_length is None:
wait_event_queue_length = 10
self._force_device_scalars = force_device_scalars
self._force_device_scalars = True
# Subclasses might still be using the old
# "force_devices_scalars: bool = False" interface, in which case we need
# to explicitly pass force_device_scalars=True in clone()
self._passed_force_device_scalars = force_device_scalars is not None
self._wait_event_queue_length = wait_event_queue_length
self._kernel_name_to_wait_event_queue: Dict[str, List[cl.Event]] = {}
self._kernel_name_to_wait_event_queue: dict[str, list[cl.Event]] = {}
if queue.device.type & cl.device_type.GPU:
if allocator is None:
warn("PyOpenCLArrayContext created without an allocator on a GPU. "
"This can lead to high numbers of memory allocations. "
"Please consider using a pyopencl.tools.MemoryPool. "
"Run with allocator=False to disable this warning.")
"Run with allocator=False to disable this warning.",
stacklevel=2)
if __debug__:
# Use "running on GPU" as a proxy for "they care about speed".
warn("You are using the PyOpenCLArrayContext on a GPU, but you "
"are running Python in debug mode. Use 'python -O' for "
"a noticeable speed improvement.")
"a noticeable speed improvement.",
stacklevel=2)
self._loopy_transform_cache: \
Dict["lp.TranslationUnit", "lp.TranslationUnit"] = {}
dict[lp.TranslationUnit, lp.TranslationUnit] = {}
self.array_types = (cla.Array,)
# TODO: Ideally this should only be `(TaggableCLArray,)`, but
# that would break the logic in the downstream users.
self.array_types = (cl_array.Array,)
def _get_fake_numpy_namespace(self):
from arraycontext.impl.pyopencl.fake_numpy import PyOpenCLFakeNumpyNamespace
return PyOpenCLFakeNumpyNamespace(self)
def _rec_map_container(
self, func: Callable[[Array], Array], array: ArrayOrContainer,
allowed_types: tuple[type, ...] | None = None, *,
default_scalar: ScalarLike | None = None,
strict: bool = False) -> ArrayOrContainer:
import arraycontext.impl.pyopencl.taggable_cl_array as tga
if allowed_types is None:
# TODO: replace with 'self.array_types' once `cla.Array` support
# is completely removed
allowed_types = (tga.TaggableCLArray,)
def _wrapper(ary):
if isinstance(ary, allowed_types):
return func(ary)
elif not strict and isinstance(ary, self.array_types):
from warnings import warn
warn(f"Invoking {type(self).__name__}.{func.__name__[1:]} with "
f"{type(ary).__name__} will be unsupported in 2023. Use "
"'to_tagged_cl_array' to convert instances to TaggableCLArray.",
DeprecationWarning, stacklevel=2)
return func(tga.to_tagged_cl_array(ary))
elif np.isscalar(ary):
if default_scalar is None:
return ary
else:
return np.array(ary).dtype.type(default_scalar)
else:
raise TypeError(
f"{type(self).__name__}.{func.__name__[1:]} invoked with "
f"an unsupported array type: got '{type(ary).__name__}', "
f"but expected one of {allowed_types}")
return rec_map_array_container(_wrapper, array)
# {{{ ArrayContext interface
def empty(self, shape, dtype):
import pyopencl.array as cl_array
return cl_array.empty(self.queue, shape=shape, dtype=dtype,
allocator=self.allocator)
def from_numpy(self, array):
import arraycontext.impl.pyopencl.taggable_cl_array as tga
def zeros(self, shape, dtype):
import pyopencl.array as cl_array
return cl_array.zeros(self.queue, shape=shape, dtype=dtype,
allocator=self.allocator)
def _from_numpy(ary):
return tga.to_device(self.queue, ary, allocator=self.allocator)
def from_numpy(self, array: np.ndarray):
import pyopencl.array as cl_array
return cl_array.to_device(self.queue, array, allocator=self.allocator)
return with_array_context(
self._rec_map_container(_from_numpy, array, (np.ndarray,), strict=True),
actx=self)
def to_numpy(self, array):
if not self._force_device_scalars and np.isscalar(array):
return array
def _to_numpy(ary):
return ary.get(queue=self.queue)
return with_array_context(
self._rec_map_container(_to_numpy, array),
actx=None)
def freeze(self, array):
def _freeze(ary):
ary.finish()
return ary.with_queue(None)
return with_array_context(self._rec_map_container(_freeze, array), actx=None)
def thaw(self, array):
def _thaw(ary):
return ary.with_queue(self.queue)
return with_array_context(self._rec_map_container(_thaw, array), actx=self)
return array.get(queue=self.queue)
def tag(self, tags: ToTagSetConvertible, array):
def _tag(ary):
return ary.tagged(tags)
return self._rec_map_container(_tag, array)
def tag_axis(self, iaxis: int, tags: ToTagSetConvertible, array):
def _tag_axis(ary):
return ary.with_tagged_axis(iaxis, tags)
return self._rec_map_container(_tag_axis, array)
def call_loopy(self, t_unit, **kwargs):
try:
t_unit = self._loopy_transform_cache[t_unit]
executor = self._loopy_transform_cache[t_unit]
except KeyError:
orig_t_unit = t_unit
t_unit = self.transform_loopy_program(t_unit)
self._loopy_transform_cache[orig_t_unit] = t_unit
executor = self.transform_loopy_program(t_unit).executor(self.context)
self._loopy_transform_cache[orig_t_unit] = executor
del orig_t_unit
evt, result = t_unit(self.queue, **kwargs, allocator=self.allocator)
evt, result = executor(self.queue, **kwargs, allocator=self.allocator)
if self._wait_event_queue_length is not False:
prg_name = t_unit.default_entrypoint.name
prg_name = executor.t_unit.default_entrypoint.name
wait_event_queue = self._kernel_name_to_wait_event_queue.setdefault(
prg_name, [])
......@@ -186,27 +261,37 @@ class PyOpenCLArrayContext(ArrayContext):
if len(wait_event_queue) > self._wait_event_queue_length:
wait_event_queue.pop(0).wait()
return result
import arraycontext.impl.pyopencl.taggable_cl_array as tga
def freeze(self, array):
array.finish()
return array.with_queue(None)
# FIXME: Inherit loopy tags for these arrays
return {name: tga.to_tagged_cl_array(ary) for name, ary in result.items()}
def thaw(self, array):
return array.with_queue(self.queue)
def clone(self):
if self._passed_force_device_scalars:
return type(self)(self.queue, self.allocator,
wait_event_queue_length=self._wait_event_queue_length,
force_device_scalars=True)
else:
return type(self)(self.queue, self.allocator,
wait_event_queue_length=self._wait_event_queue_length)
# }}}
def transform_loopy_program(self, t_unit):
# {{{ transform_loopy_program
def transform_loopy_program(self, t_unit: lp.TranslationUnit) -> lp.TranslationUnit:
from warnings import warn
warn("Using arraycontext.PyOpenCLArrayContext.transform_loopy_program "
"to transform a program. This is deprecated and will stop working "
"in 2022. Instead, subclass PyOpenCLArrayContext and implement "
"the specific logic required to transform the program for your "
"package or application. Check higher-level packages "
warn("Using the base "
f"{type(self).__name__}.transform_loopy_program "
"to transform a translation unit. "
"This is largely a no-op and unlikely to result in fast generated "
"code."
f"Instead, subclass {type(self).__name__} and implement "
"the specific transform logic required to transform the program "
"for your package or application. Check higher-level packages "
"(e.g. meshmode), which may already have subclasses you may want "
"to build on.",
DeprecationWarning, stacklevel=2)
UntransformedCodeWarning, stacklevel=2)
# accommodate loopy with and without kernel callables
......@@ -220,40 +305,17 @@ class PyOpenCLArrayContext(ArrayContext):
"to create this kernel?")
all_inames = default_entrypoint.all_inames()
# FIXME: This could be much smarter.
inner_iname = None
# import with underscore to avoid DeprecationWarning
from arraycontext.metadata import _FirstAxisIsElementsTag
if (len(default_entrypoint.instructions) == 1
and isinstance(default_entrypoint.instructions[0], lp.Assignment)
and any(isinstance(tag, _FirstAxisIsElementsTag)
# FIXME: Firedrake branch lacks kernel tags
for tag in getattr(default_entrypoint, "tags", ()))):
stmt, = default_entrypoint.instructions
out_inames = [v.name for v in stmt.assignee.index_tuple]
assert out_inames
outer_iname = out_inames[0]
if len(out_inames) >= 2:
inner_iname = out_inames[1]
elif "iel" in all_inames:
outer_iname = "iel"
inner_iname = None
if "idof" in all_inames:
inner_iname = "idof"
elif "i0" in all_inames:
if "i0" in all_inames:
outer_iname = "i0"
if "i1" in all_inames:
inner_iname = "i1"
else:
raise RuntimeError(
"Unable to reason what outer_iname and inner_iname "
f"needs to be; all_inames is given as: {all_inames}"
)
return t_unit
if inner_iname is not None:
t_unit = lp.split_iname(t_unit, inner_iname, 16, inner_tag="l.0")
......@@ -261,23 +323,24 @@ class PyOpenCLArrayContext(ArrayContext):
return t_unit
def tag(self, tags: Union[Sequence[Tag], Tag], array):
# Sorry, not capable.
return array
def tag_axis(self, iaxis, tags: Union[Sequence[Tag], Tag], array):
# Sorry, not capable.
return array
# }}}
def clone(self):
return type(self)(self.queue, self.allocator,
wait_event_queue_length=self._wait_event_queue_length,
force_device_scalars=self._force_device_scalars)
# {{{ properties
@property
def permits_inplace_modification(self):
return True
@property
def supports_nonscalar_broadcasting(self):
return False
@property
def permits_advanced_indexing(self):
return False
# }}}
# }}}
# vim: foldmethod=marker