OUTDATED DOCUMENTATION

You are viewing archived documentation from the old Numba documentation site. The current documentation is located at https://numba.readthedocs.io.

A Map of the Numba Repository

The Numba repository is quite large, and due to age has functionality spread around many locations. To help orient developers, this document will try to summarize where different categories of functionality can be found.

Note

It is likely that the organization of the code base will change in the future to improve organization. Follow issue #3807 for more details.

Support Files

Build and Packaging

• setup.py - Standard Python distutils/setuptools script

• MANIFEST.in - Distutils packaging instructions

• requirements.txt - Pip package requirements, not used by conda

• versioneer.py - Handles automatic setting of version in installed package from git tags

• .flake8 - Preferences for code formatting. Files should be fixed and removed from the exception list as time allows.

• .pre-commit-config.yaml - Configuration file for pre-commit hooks.

• buildscripts/condarecipe.local - Conda build recipe

• buildscripts/condarecipe_clone_icc_rt - Recipe to build a standalone icc_rt package.

Continuous Integration

• azure-pipelines.yml - Azure Pipelines CI config (active: Win/Mac/Linux)

• buildscripts/azure/ - Azure Pipeline configuration for specific platforms

• buildscripts/appveyor/ - Appveyor build scripts

• buildscripts/incremental/ - Generic scripts for building Numba on various CI systems

• codecov.yml - Codecov.io coverage reporting

Documentation

• LICENSE - License for Numba

• LICENSES.third-party - License for third party code vendored into Numba

• README.rst - README for repo, also uploaded to PyPI

• CONTRIBUTING.md - Documentation on how to contribute to project (out of date, should be updated to point to Sphinx docs)

• CHANGE_LOG - History of Numba releases, also directly embedded into Sphinx documentation

• docs/ - Documentation source

• docs/_templates/ - Directory for templates (to override defaults with Sphinx theme)

• docs/Makefile - Used to build Sphinx docs with make

• docs/source - ReST source for Numba documentation

• docs/_static/ - Static CSS and image assets for Numba docs

• docs/gh-pages.py - Utility script to update Numba docs (stored as gh-pages)

• docs/make.bat - Not used (remove?)

• numba/scripts/generate_lower_listing.py - Dump all registered implementations decorated with @lower* for reference documentation. Currently misses implementations from the higher level extension API.

Numba Source Code

Numba ships with both the source code and tests in one package.

• numba/ - all of the source code and tests

Public API

These define aspects of the public Numba interface.

• numba/core/decorators.py - User-facing decorators for compiling regular functions on the CPU

• numba/core/extending.py - Public decorators for extending Numba (overload, intrinsic, etc)

• numba/core/ccallback.py - @cfunc decorator for compiling functions to a fixed C signature. Used to make callbacks.

• numba/np/ufunc/decorators.py - ufunc/gufunc compilation decorators

• numba/core/config.py - Numba global config options and environment variable handling

• numba/core/annotations - Gathering and printing type annotations of Numba IR

• numba/core/annotations/pretty_annotate.py - Code highlighting of Numba functions and types (both ANSI terminal and HTML)

Dispatching

• numba/core/dispatcher.py - Dispatcher objects are compiled functions produced by @jit. A dispatcher has different implementations for different type signatures.

• numba/_dispatcher.h | numba/_dispatcher.c - C interface to C++ dispatcher implementation

• numba/_dispatcherimpl.cpp - C++ dispatcher implementation (for speed on common data types)

Compiler Pipeline

• numba/core/compiler.py - Compiler pipelines and flags

• numba/core/errors.py - Numba exception and warning classes

• numba/core/ir.py - Numba IR data structure objects

• numba/core/bytecode.py - Bytecode parsing and function identity (??)

• numba/core/interpreter.py - Translate Python interpreter bytecode to Numba IR

• numba/core/analysis.py - Utility functions to analyze Numba IR (variable lifetime, prune branches, etc)

• numba/core/dataflow.py - Dataflow analysis for Python bytecode (used in analysis.py)

• numba/core/controlflow.py - Control flow analysis of Numba IR and Python bytecode

• numba/core/typeinfer.py - Type inference algorithm

• numba/core/transforms.py - Numba IR transformations

• numba/core/rewrites - Rewrite passes used by compiler

• numba/core/rewrites/__init__.py - Loads all rewrite passes so they are put into the registry

• numba/core/rewrites/registry.py - Registry object for collecting rewrite passes

• numba/core/rewrites/ir_print.py - Write print() calls into special print nodes in the IR

• numba/core/rewrites/static_raise.py - Converts exceptions with static arguments into a special form that can be lowered

• numba/core/rewrites/macros.py - Generic support for macro expansion in the Numba IR

• numba/core/rewrites/static_getitem.py - Rewrites getitem and setitem with constant arguments to allow type inference

• numba/core/rewrites/static_binop.py - Rewrites binary operations (specifically **) with constant arguments so faster code can be generated

• numba/core/inline_closurecall.py - Inlines body of closure functions to call site. Support for array comprehensions, reduction inlining, and stencil inlining.

• numba/core/postproc.py - Postprocessor for Numba IR that computes variable lifetime, inserts del operations, and handles generators

• numba/core/lowering.py - General implementation of lowering Numba IR to LLVM

• numba/core/withcontexts.py - General scaffolding for implementing context managers in nopython mode, and the objectmode context manager

• numba/core/pylowering.py - Lowering of Numba IR in object mode

• numba/core/pythonapi.py - LLVM IR code generation to interface with CPython API

Type Management

• numba/core/typeconv/ - Implementation of type casting and type signature matching in both C++ and Python

• numba/capsulethunk.h - Used by typeconv

• numba/core/types/ - definition of the Numba type hierarchy, used everywhere in compiler to select implementations

• numba/core/consts.py - Constant inference (used to make constant values available during codegen when possible)

• numba/core/datamodel - LLVM IR representations of data types in different contexts

• numba/core/datamodel/models.py - Models for most standard types

• numba/core/datamodel/registry.py - Decorator to register new data models

• numba/core/datamodel/packer.py - Pack typed values into a data structure

• numba/core/datamodel/testing.py - Data model tests (this should move??)

• numba/core/datamodel/manager.py - Map types to data models

Compiled Extensions

Numba uses a small amount of compiled C/C++ code for core functionality, like dispatching and type matching where performance matters, and it is more convenient to encapsulate direct interaction with CPython APIs.

• numba/_arraystruct.h - Struct for holding NumPy array attributes. Used in helperlib and the Numba Runtime.

• numba/_helperlib.c - C functions required by Numba compiled code at runtime. Linked into ahead-of-time compiled modules

• numba/_helpermod.c - Python extension module with pointers to functions from _helperlib.c and _npymath_exports.c

• numba/_npymath_exports.c - Export function pointer table to NumPy C math functions

• numba/_dynfuncmod.c - Python extension module exporting _dynfunc.c functionality

• numba/_dynfunc.c - C level Environment and Closure objects (keep in sync with numba/target/base.py)

• numba/mathnames.h - Macros for defining names of math functions

• numba/_pymodule.h - C macros for Python 2/3 portable naming of C API functions

• numba/mviewbuf.c - Handles Python memoryviews

• numba/_typeof.h | numba/_typeof.c - C implementation of type fingerprinting, used by dispatcher

• numba/_numba_common.h - Portable C macro for marking symbols that can be shared between object files, but not outside the library.

Misc Support

• numba/_version.py - Updated by versioneer

• numba/core/runtime - Language runtime. Currently manages reference-counted memory allocated on the heap by Numba-compiled functions

• numba/core/ir_utils.py - Utility functions for working with Numba IR data structures

• numba/core/cgutils.py - Utility functions for generating common code patterns in LLVM IR

• numba/core/utils.py - Python 2 backports of Python 3 functionality (also imports local copy of six)

• numba/misc/appdirs.py - Vendored package for determining application config directories on every platform

• numba/core/compiler_lock.py - Global compiler lock because Numba’s usage of LLVM is not thread-safe

• numba/misc/special.py - Python stub implementations of special Numba functions (prange, gdb*)

• numba/core/itanium_mangler.py - Python implementation of Itanium C++ name mangling

• numba/misc/findlib.py - Helper function for locating shared libraries on all platforms

• numba/core/debuginfo.py - Helper functions to construct LLVM IR debug info

• numba/core/unsafe/refcount.py - Read reference count of object

• numba/core/unsafe/eh.py - Exception handling helpers

• numba/core/unsafe/nrt.py - Numba runtime (NRT) helpers

• numba/cpython/unsafe/tuple.py - Replace a value in a tuple slot

• numba/np/unsafe/ndarray.py - NumPy array helpers

• numba/core/unsafe/bytes.py - Copying and dereferencing data from void pointers

• numba/misc/dummyarray.py - Used by GPU backends to hold array information on the host, but not the data.

• numba/core/callwrapper.py - Handles argument unboxing and releasing the GIL when moving from Python to nopython mode

• numba/np/numpy_support.py - Helper functions for working with NumPy and translating Numba types to and from NumPy dtypes.

• numba/core/tracing.py - Decorator for tracing Python calls and emitting log messages

• numba/core/funcdesc.py - Classes for describing function metadata (used in the compiler)

• numba/core/sigutils.py - Helper functions for parsing and normalizing Numba type signatures

• numba/core/serialize.py - Support for pickling compiled functions

• numba/core/caching.py - Disk cache for compiled functions

• numba/np/npdatetime.py - Helper functions for implementing NumPy datetime64 support

Core Python Data Types

• numba/_hashtable.h | numba/_hashtable.c - Adaptation of the Python 3.7 hash table implementation

• numba/cext/dictobject.h | numba/cext/dictobject.c - C level implementation of typed dictionary

• numba/typed/dictobject.py - Nopython mode wrapper for typed dictionary

• numba/cext/listobject.h | numba/cext/listobject.c - C level implementation of typed list

• numba/typed/listobject.py - Nopython mode wrapper for typed list

• numba/typed/typedobjectutils.py - Common utilities for typed dictionary and list

• numba/cpython/unicode.py - Unicode strings (Python 3.5 and later)

• numba/typed - Python interfaces to statically typed containers

• numba/typed/typeddict.py - Python interface to typed dictionary

• numba/typed/typedlist.py - Python interface to typed list

• numba/experimental/jitclass - Implementation of experimental JIT compilation of Python classes

• numba/core/generators.py - Support for lowering Python generators

Math

• numba/_random.c - Reimplementation of NumPy / CPython random number generator

• numba/_lapack.c - Wrappers for calling BLAS and LAPACK functions (requires SciPy)

ParallelAccelerator

Code transformation passes that extract parallelizable code from a function and convert it into multithreaded gufunc calls.

• numba/parfors/parfor.py - General ParallelAccelerator

• numba/parfors/parfor_lowering.py - gufunc lowering for ParallelAccelerator

• numba/parfors/array_analysis.py - Array analysis passes used in ParallelAccelerator

Stencil

Implementation of @stencil:

• numba/stencils/stencil.py - Stencil function decorator (implemented without ParallelAccelerator)

• numba/stencils/stencilparfor.py - ParallelAccelerator implementation of stencil

Debugging Support

• numba/misc/gdb_hook.py - Hooks to jump into GDB from nopython mode

• numba/misc/cmdlang.gdb - Commands to setup GDB for setting explicit breakpoints from Python

Type Signatures (CPU)

Some (usually older) Numba supported functionality separates the declaration of allowed type signatures from the definition of implementations. This package contains registries of type signatures that must be matched during type inference.

• numba/core/typing - Type signature module

• numba/core/typing/templates.py - Base classes for type signature templates

• numba/core/typing/cmathdecl.py - Python complex math (cmath) module

• numba/core/typing/bufproto.py - Interpreting objects supporting the buffer protocol

• numba/core/typing/mathdecl.py - Python math module

• numba/core/typing/listdecl.py - Python lists

• numba/core/typing/builtins.py - Python builtin global functions and operators

• numba/core/typing/randomdecl.py - Python and NumPy random modules

• numba/core/typing/setdecl.py - Python sets

• numba/core/typing/npydecl.py - NumPy ndarray (and operators), NumPy functions

• numba/core/typing/arraydecl.py - Python array module

• numba/core/typing/context.py - Implementation of typing context (class that collects methods used in type inference)

• numba/core/typing/collections.py - Generic container operations and namedtuples

• numba/core/typing/ctypes_utils.py - Typing ctypes-wrapped function pointers

• numba/core/typing/enumdecl.py - Enum types

• numba/core/typing/cffi_utils.py - Typing of CFFI objects

• numba/core/typing/typeof.py - Implementation of typeof operations (maps Python object to Numba type)

• numba/core/typing/npdatetime.py - Datetime dtype support for NumPy arrays

Target Implementations (CPU)

Implementations of Python / NumPy functions and some data models. These modules are responsible for generating LLVM IR during lowering. Note that some of these modules do not have counterparts in the typing package because newer Numba extension APIs (like overload) allow typing and implementation to be specified together.

• numba/core/cpu.py - Context for code gen on CPU

• numba/core/base.py - Base class for all target contexts

• numba/core/codegen.py - Driver for code generation

• numba/core/boxing.py - Boxing and unboxing for most data types

• numba/core/intrinsics.py - Utilities for converting LLVM intrinsics to other math calls

• numba/core/callconv.py - Implements different calling conventions for Numba-compiled functions

• numba/core/options.py - Container for options that control lowering

• numba/core/optional.py - Special type representing value or None

• numba/core/registry.py - Registry object for collecting implementations for a specific target

• numba/core/imputils.py - Helper functions for lowering

• numba/core/externals.py - Registers external C functions needed to link generated code

• numba/core/fastmathpass.py - Rewrite pass to add fastmath attributes to function call sites and binary operations

• numba/core/removerefctpass.py - Rewrite pass to remove unnecessary incref/decref pairs

• numba/core/descriptors.py - empty base class for all target descriptors (is this needed?)

• numba/cpython/builtins.py - Python builtin functions and operators

• numba/cpython/cmathimpl.py - Python complex math module

• numba/cpython/enumimpl.py - Enum objects

• numba/cpython/hashing.py - Hashing algorithms

• numba/cpython/heapq.py - Python heapq module

• numba/cpython/iterators.py - Iterable data types and iterators

• numba/cpython/listobj.py - Python lists

• numba/cpython/mathimpl.py - Python math module

• numba/cpython/numbers.py - Numeric values (int, float, etc)

• numba/cpython/printimpl.py - Print function

• numba/cpython/randomimpl.py - Python and NumPy random modules

• numba/cpython/rangeobj.py - Python range objects

• numba/cpython/slicing.py - Slice objects, and index calculations used in slicing

• numba/cpython/setobj.py - Python set type

• numba/cpython/tupleobj.py - Tuples (statically typed as immutable struct)

• numba/misc/cffiimpl.py - CFFI functions

• numba/misc/quicksort.py - Quicksort implementation used with list and array objects

• numba/misc/mergesort.py - Mergesort implementation used with array objects

• numba/np/arraymath.py - Math operations on arrays (both Python and NumPy)

• numba/np/arrayobj.py - Array operations (both NumPy and buffer protocol)

• numba/np/linalg.py - NumPy linear algebra operations

• numba/np/npdatetime.py - NumPy datetime operations

• numba/np/npyfuncs.py - Kernels used in generating some NumPy ufuncs

• numba/np/npyimpl.py - Implementations of most NumPy ufuncs

• numba/np/polynomial.py - numpy.roots function

• numba/np/ufunc_db.py - Big table mapping types to ufunc implementations

Ufunc Compiler and Runtime

• numba/np/ufunc - ufunc compiler implementation

• numba/np/ufunc/_internal.h | numba/np/ufunc/_internal.c - Python extension module with helper functions that use CPython & NumPy C API

• numba/np/ufunc/_ufunc.c - Used by _internal.c

• numba/np/ufunc/deviceufunc.py - Custom ufunc dispatch for non-CPU targets

• numba/np/ufunc/gufunc_scheduler.h | numba/np/ufunc/gufunc_scheduler.cpp - Schedule work chunks to threads

• numba/np/ufunc/dufunc.py - Special ufunc that can compile new implementations at call time

• numba/np/ufunc/ufuncbuilder.py - Top-level orchestration of ufunc/gufunc compiler pipeline

• numba/np/ufunc/sigparse.py - Parser for generalized ufunc indexing signatures

• numba/np/ufunc/parallel.py - Codegen for parallel target

• numba/np/ufunc/array_exprs.py - Rewrite pass for turning array expressions in regular functions into ufuncs

• numba/np/ufunc/wrappers.py - Wrap scalar function kernel with loops

• numba/np/ufunc/workqueue.h | numba/np/ufunc/workqueue.c - Threading backend based on pthreads/Windows threads and queues

• numba/np/ufunc/omppool.cpp - Threading backend based on OpenMP

• numba/np/ufunc/tbbpool.cpp - Threading backend based on TBB

Unit Tests (CPU)

CPU unit tests (GPU target unit tests listed in later sections

• runtests.py - Convenience script that launches test runner and turns on full compiler tracebacks

• run_coverage.py - Runs test suite with coverage tracking enabled

• .coveragerc - Coverage.py configuration

• numba/runtests.py - Entry point to unittest runner

• numba/testing/_runtests.py - Implementation of custom test runner command line interface

• numba/tests/test_* - Test cases

• numba/tests/*_usecases.py - Python functions compiled by some unit tests

• numba/tests/support.py - Helper functions for testing and special TestCase implementation

• numba/tests/dummy_module.py - Module used in test_dispatcher.py

• numba/tests/npyufunc - ufunc / gufunc compiler tests

• numba/testing - Support code for testing

• numba/testing/loader.py - Find tests on disk

• numba/testing/notebook.py - Support for testing notebooks

• numba/testing/main.py - Numba test runner

Command Line Utilities

• bin/numba - Command line stub, delegates to main in numba_entry.py

• numba/misc/numba_entry.py - Main function for numba command line tool

• numba/pycc - Ahead of time compilation of functions to shared library extension

• numba/pycc/__init__.py - Main function for pycc command line tool

• numba/pycc/cc.py - User-facing API for tagging functions to compile ahead of time

• numba/pycc/compiler.py - Compiler pipeline for creating standalone Python extension modules

• numba/pycc/llvm_types.py - Aliases to LLVM data types used by compiler.py

• numba/pycc/pycc - Stub to call main function. Is this still used?

• numba/pycc/modulemixin.c - C file compiled into every compiled extension. Pulls in C source from Numba core that is needed to make extension standalone

• numba/pycc/platform.py - Portable interface to platform-specific compiler toolchains

• numba/pycc/decorators.py - Deprecated decorators for tagging functions to compile. Use cc.py instead.

CUDA GPU Target

Note that the CUDA target does reuse some parts of the CPU target.

• numba/cuda/ - The implementation of the CUDA (NVIDIA GPU) target and associated unit tests

• numba/cuda/decorators.py - Compiler decorators for CUDA kernels and device functions

• numba/cuda/dispatcher.py - Dispatcher for CUDA JIT functions

• numba/cuda/printimpl.py - Special implementation of device printing

• numba/cuda/libdevice.py - Registers libdevice functions

• numba/cuda/kernels/ - Custom kernels for reduction and transpose

• numba/cuda/device_init.py - Initializes the CUDA target when imported

• numba/cuda/compiler.py - Compiler pipeline for CUDA target

• numba/cuda/intrinsic_wrapper.py - CUDA device intrinsics (shuffle, ballot, etc)

• numba/cuda/initialize.py - Defered initialization of the CUDA device and subsystem. Called only when user imports numba.cuda

• numba/cuda/simulator_init.py - Initalizes the CUDA simulator subsystem (only when user requests it with env var)

• numba/cuda/random.py - Implementation of random number generator

• numba/cuda/api.py - User facing APIs imported into numba.cuda.*

• numba/cuda/stubs.py - Python placeholders for functions that only can be used in GPU device code

• numba/cuda/simulator/ - Simulate execution of CUDA kernels in Python interpreter

• numba/cuda/vectorizers.py - Subclasses of ufunc/gufunc compilers for CUDA

• numba/cuda/args.py - Management of kernel arguments, including host<->device transfers

• numba/cuda/target.py - Typing and target contexts for GPU

• numba/cuda/cudamath.py - Type signatures for math functions in CUDA Python

• numba/cuda/errors.py - Validation of kernel launch configuration

• numba/cuda/nvvmutils.py - Helper functions for generating NVVM-specific IR

• numba/cuda/testing.py - Support code for creating CUDA unit tests and capturing standard out

• numba/cuda/cudadecl.py - Type signatures of CUDA API (threadIdx, blockIdx, atomics) in Python on GPU

• numba/cuda/cudaimpl.py - Implementations of CUDA API functions on GPU

• numba/cuda/codegen.py - Code generator object for CUDA target

• numba/cuda/cudadrv/ - Wrapper around CUDA driver API

• numba/cuda/tests/ - CUDA unit tests, skipped when CUDA is not detected

• numba/cuda/tests/cudasim/ - Tests of CUDA simulator

• numba/cuda/tests/nocuda/ - Tests for NVVM functionality when CUDA not present

• numba/cuda/tests/cudapy/ - Tests of compiling Python functions for GPU

• numba/cuda/tests/cudadrv/ - Tests of Python wrapper around CUDA API

ROCm GPU Target

Note that the ROCm target does reuse some parts of the CPU target, and duplicates some code from CUDA target. A future refactoring could pull out the common subset of CUDA and ROCm. An older version of this target was based on the HSA API, so “hsa” appears in many places.

• numba/roc - ROCm GPU target for AMD GPUs

• numba/roc/descriptor.py - TargetDescriptor subclass for ROCm target

• numba/roc/enums.py - Internal constants

• numba/roc/mathdecl.py - Declarations of math functions that can be used on device

• numba/roc/mathimpl.py - Implementations of math functions for device

• numba/roc/compiler.py - Compiler pipeline for ROCm target

• numba/roc/hlc - Wrapper around LLVM interface for AMD GPU

• numba/roc/initialize.py - Register ROCm target for ufunc/gufunc compiler

• numba/roc/hsadecl.py - Type signatures for ROCm device API in Python

• numba/roc/hsaimpl.py - Implementations of ROCm device API

• numba/roc/dispatch.py - ufunc/gufunc dispatcher

• numba/roc/README.md - Notes on testing target (should be deleted)

• numba/roc/api.py - Host API for ROCm actions

• numba/roc/gcn_occupancy.py - Heuristic to compute occupancy of kernels

• numba/roc/stubs.py - Host stubs for device functions

• numba/roc/vectorizers.py - Builds ufuncs

• numba/roc/target.py - Target and typing contexts

• numba/roc/hsadrv - Python wrapper around ROCm (based on HSA) driver API calls

• numba/roc/codegen.py - Codegen subclass for ROCm target

• numba/roc/decorators.py - @jit decorator for kernels and device functions

• numba/roc/servicelib/threadlocal.py - Thread-local stack used by ROC targets

• numba/roc/servicelib/service.py - Should be removed?

• numba/roc/tests - Unit tests for ROCm target

• numba/roc/tests/hsapy - Tests of compiling ROCm kernels written in Python syntax

• numba/roc/tests/hsadrv - Tests of Python wrapper on platform API.