Intel® oneAPI Threading Building Blocks (oneTBB)

2022.3.1 · active · verified Sat Apr 11

The `tbb` library provides Intel® oneAPI Threading Building Blocks (oneTBB), a C++ template library for parallel programming. The Python package primarily distributes the oneTBB C++ runtime libraries and headers, enabling other Python packages (e.g., NumPy, SciPy) to leverage TBB for parallelization. It is currently at version 2022.3.1 and sees regular updates, typically a few times per year, to introduce new features and address issues.

Warnings

gotcha The `tbb` Python package primarily provides Intel oneTBB C++ runtime libraries and headers. It does not offer direct Python APIs for parallelizing Python code. Users typically interact with libraries like NumPy, SciPy, or scikit-learn, which may be *built against* TBB for their internal parallelization.
Fix: Understand that `tbb` is a backend dependency provider. For Python parallelization, consider libraries like `joblib`, `numba`, or native `multiprocessing`.
breaking Environment variable names for C++ include paths changed in oneTBB 2022.2.0. `CPATH` was replaced by `C_INCLUDE_PATH` and `CPLUS_INCLUDE_PATH` to prevent unintended compiler warnings.
Fix: Update C++ build scripts or environment configurations to use `C_INCLUDE_PATH` and `CPLUS_INCLUDE_PATH` instead of `CPATH` when building against TBB 2022.2.0 or newer.
gotcha While the `tbb` Python package provides the necessary shared objects and headers, direct C++ development or Python extensions that link against TBB still require a C++ compiler and appropriate build system setup.
Fix: Ensure your development environment has a compatible C++ compiler (e.g., GCC, Clang, MSVC) and a build system (e.g., CMake, Make) configured to locate and link against the TBB libraries.
breaking The primary C++ namespace for TBB migrated from `tbb` to `oneapi::tbb` with the introduction of oneAPI. While aliases often exist for backward compatibility, new C++ code should use the `oneapi::tbb` namespace.
Fix: When writing or updating C++ code that uses TBB, prefer `oneapi::tbb::` for types and functions. For older code, verify that the TBB version you are linking against provides the necessary `tbb::` aliases.
deprecated The `tbb::task_scheduler_init` class has been largely deprecated in modern TBB versions in favor of `tbb::task_arena` for finer-grained control over task execution and thread management.
Fix: For C++ code, migrate usage from `tbb::task_scheduler_init` to `tbb::task_arena` to leverage improved control and modern best practices for task scheduling.

Install

pip install tbb Install the TBB Python package

Imports

version_info
```
from tbb import version_info
```
Used to retrieve the installed TBB C++ library version.
include_dir
```
from tbb import include_dir
```
Used to find the path to TBB C++ headers for compilation.
lib_dir
```
from tbb import lib_dir
```
Used to find the path to TBB C++ shared libraries for linking.

Quickstart

Demonstrates how to import the `tbb` package to retrieve information about the installed Intel oneTBB C++ library, such as its version and file system paths. This is useful for verifying installation or for C++/Python interoperability projects.

import tbb
import os

# The 'tbb' Python package primarily provides access to the underlying
# C++ library paths and version, not direct Python parallel computing APIs.
# Other Python libraries (e.g., NumPy, SciPy) might be built to use TBB
# and will detect it automatically if installed.

print(f"TBB package version: {tbb.version_info}")

# Paths useful for C++ projects or Python extensions that link against TBB
print(f"TBB C++ include directory: {tbb.include_dir()}")
print(f"TBB C++ library directory: {tbb.lib_dir()}")

# Example of checking if TBB-enabled features are available in another library
# (Conceptual, 'some_tbb_enabled_lib' is a placeholder)
# try:
#     import some_tbb_enabled_lib
#     if hasattr(some_tbb_enabled_lib, 'is_tbb_enabled') and some_tbb_enabled_lib.is_tbb_enabled():
#         print("A TBB-enabled Python library is detected and can use TBB.")
# except ImportError:
#     print("No specific TBB-enabled Python library found for direct usage example.")

view raw JSON →