threadpoolctl 1.1.0

threadpoolctl

Thread-pool Controls

Python helpers to limit the number of threads used in the threadpool-backed of common native libraries used for scientific computing and data science (e.g. BLAS and OpenMP).

Fine control of the underlying thread-pool size can be useful in workloads that involve nested parallelism so as to mitigate oversubscription issues.

Installation

• For users, install the last published version from PyPI:

  pip install threadpoolctl
  

• For contributors, install from the source repository in developer mode:

  pip install -r dev-requirements.txt
  flit install --symlink
  

  then you run the tests with pytest:

  pytest
  

Usage

Runtime Introspection

Introspect the current state of the threadpool-enabled runtime libraries that are loaded when importing Python packages:

>>> from threadpoolctl import threadpool_info
>>> from pprint import pprint
>>> pprint(threadpool_info())
[]

>>> import numpy
>>> pprint(threadpool_info())
[{'filepath': '/opt/venvs/py37/lib/python3.7/site-packages/numpy/.libs/libopenblasp-r0-382c8f3a.3.5.dev.so',
  'internal_api': 'openblas',
  'num_threads': 4,
  'prefix': 'libopenblas',
  'user_api': 'blas',
  'version': '0.3.5.dev'}]

>>> import xgboost
>>> pprint(threadpool_info())
[{'filepath': '/opt/venvs/py37/lib/python3.7/site-packages/numpy/.libs/libopenblasp-r0-382c8f3a.3.5.dev.so',
  'internal_api': 'openblas',
  'num_threads': 4,
  'prefix': 'libopenblas',
  'user_api': 'blas',
  'version': '0.3.5.dev'},
 {'filepath': '/opt/venvs/py37/lib/python3.7/site-packages/scipy/.libs/libopenblasp-r0-8dca6697.3.0.dev.so',
  'internal_api': 'openblas',
  'num_threads': 4,
  'prefix': 'libopenblas',
  'user_api': 'blas',
  'version': None},
 {'filepath': '/usr/lib/x86_64-linux-gnu/libgomp.so.1',
  'internal_api': 'openmp',
  'num_threads': 4,
  'prefix': 'libgomp',
  'user_api': 'openmp',
  'version': None}]

Set the maximum size of thread-pools

Control the number of threads used by the underlying runtime libraries in specific sections of your Python program:

from threadpoolctl import threadpool_limits
import numpy as np


with threadpool_limits(limits=1, user_api='blas'):
    # In this block, calls to blas implementation (like openblas or MKL)
    # will be limited to use only one thread. They can thus be used jointly
    # with thread-parallelism.
    a = np.random.randn(1000, 1000)
    a_squared = a @ a

Known limitation

threadpool_limits does not act as expected in nested parallel loops managed by distinct OpenMP runtime implementations (for instance libgomp from GCC and libomp from clang/llvm or libiomp from ICC).

See the test_openmp_nesting() function in tests/test_threadpoolctl.py for an example.

Maintainers

To make a release:

Bump the version number (__version__) in threadpoolctl.py.

Build the distribution archives:

pip install flit
flit build

Check the contents of dist/.

If everything is fine, make a commit for the release, tag it, push the tag to github and then:

flit publish