Just-in-Time compilation¶

JIT functions¶

@numba.jit(signature=None, nopython=False, nogil=False, cache=False, forceobj=False, parallel=False, error_model='python', fastmath=False, locals={}, boundscheck=False)¶

Compile the decorated function on-the-fly to produce efficient machine code. All parameters are optional.

If present, the signature is either a single signature or a list of signatures representing the expected Types and signatures of function arguments and return values. Each signature can be given in several forms:

A tuple of Types and signatures arguments (for example (numba.int32, numba.double)) representing the types of the function’s arguments; Numba will then infer an appropriate return type from the arguments.
A call signature using Types and signatures, specifying both return type and argument types. This can be given in intuitive form (for example numba.void(numba.int32, numba.double)).
A string representation of one of the above, for example "void(int32, double)". All type names used in the string are assumed to be defined in the numba.types module.

nopython and nogil are boolean flags. locals is a mapping of local variable names to Types and signatures.

This decorator has several modes of operation:

If one or more signatures are given in signature, a specialization is compiled for each of them. Calling the decorated function will then try to choose the best matching signature, and raise a TypeError if no appropriate conversion is available for the function arguments. If converting succeeds, the compiled machine code is executed with the converted arguments and the return value is converted back according to the signature.
If no signature is given, the decorated function implements lazy compilation. Each call to the decorated function will try to re-use an existing specialization if it exists (for example, a call with two integer arguments may re-use a specialization for argument types (numba.int64, numba.int64)). If no suitable specialization exists, a new specialization is compiled on-the-fly, stored for later use, and executed with the converted arguments.

If true, nopython forces the function to be compiled in nopython mode. If not possible, compilation will raise an error.

If true, forceobj forces the function to be compiled in object mode. Since object mode is slower than nopython mode, this is mostly useful for testing purposes.

If true, nogil tries to release the global interpreter lock inside the compiled function. The GIL will only be released if Numba can compile the function in nopython mode, otherwise a compilation warning will be printed.

If true, cache enables a file-based cache to shorten compilation times when the function was already compiled in a previous invocation. The cache is maintained in the __pycache__ subdirectory of the directory containing the source file; if the current user is not allowed to write to it, though, it falls back to a platform-specific user-wide cache directory (such as $HOME/.cache/numba on Unix platforms).

If true, parallel enables the automatic parallelization of a number of common Numpy constructs as well as the fusion of adjacent parallel operations to maximize cache locality.

The error_model option controls the divide-by-zero behavior. Setting it to ‘python’ causes divide-by-zero to raise exception like CPython. Setting it to ‘numpy’ causes divide-by-zero to set the result to +/-inf or nan.

Not all functions can be cached, since some functionality cannot be always persisted to disk. When a function cannot be cached, a warning is emitted.

If true, fastmath enables the use of otherwise unsafe floating point transforms as described in the LLVM documentation. Further, if Intel SVML is installed faster but less accurate versions of some math intrinsics are used (answers to within 4 ULP).

If True, boundscheck enables bounds checking for array indices. Out of bounds accesses will raise IndexError. The default is to not do bounds checking. If bounds checking is disabled, out of bounds accesses can produce garbage results or segfaults. However, enabling bounds checking will slow down typical functions, so it is recommended to only use this flag for debugging. You can also set the NUMBA_BOUNDSCHECK environment variable to 0 or 1 to globally override this flag.

The locals dictionary may be used to force the Types and signatures of particular local variables, for example if you want to force the use of single precision floats at some point. In general, we recommend you let Numba’s compiler infer the types of local variables by itself.

Here is an example with two signatures:

@jit(["int32(int32)", "float32(float32)"], nopython=True)
def f(x): ...

Not putting any parentheses after the decorator is equivalent to calling the decorator without any arguments, i.e.:

@jit
def f(x): ...

is equivalent to:

@jit()
def f(x): ...

The decorator returns a Dispatcher object.

Note

If no signature is given, compilation errors will be raised when the actual compilation occurs, i.e. when the function is first called with some given argument types.

Note

Compilation can be influenced by some dedicated Environment variables.

Generated JIT functions¶

@numba.generated_jit(nopython=False, nogil=False, cache=False, forceobj=False, locals={})¶

Like the jit() decorator, but calls the decorated function at compile-time, passing the types of the function’s arguments. The decorated function must return a callable which will be compiled as the function’s implementation for those types, allowing flexible kinds of specialization.

The generated_jit() decorator returns a Dispatcher object.

Dispatcher objects¶

class Dispatcher¶

The class of objects created by calling jit() or generated_jit(). You shouldn’t try to create such an object in any other way. Calling a Dispatcher object calls the compiled specialization for the arguments with which it is called, letting it act as an accelerated replacement for the Python function which was compiled.

In addition, Dispatcher objects have the following methods and attributes:

py_func¶: The pure Python function which was compiled.

inspect_types(file=None, pretty=False)¶: Print out a listing of the function source code annotated line-by-line with the corresponding Numba IR, and the inferred types of the various variables. If file is specified, printing is done to that file object, otherwise to sys.stdout. If pretty is set to True then colored ANSI will be produced in a terminal and HTML in a notebook.

See also

Numba architecture

inspect_llvm(signature=None)¶: Return a dictionary keying compiled function signatures to the human readable LLVM IR generated for the function. If the signature keyword is specified a string corresponding to that individual signature is returned.

inspect_asm(signature=None)¶: Return a dictionary keying compiled function signatures to the human-readable native assembly code for the function. If the signature keyword is specified a string corresponding to that individual signature is returned.

inspect_cfg(signature=None, show_wrapped)¶

Return a dictionary keying compiled function signatures to the control-flow graph objects for the function. If the signature keyword is specified a string corresponding to that individual signature is returned.

The control-flow graph objects can be stringified (str or repr) to get the textual representation of the graph in DOT format. Or, use its .display(filename=None, view=False) method to plot the graph. The filename option can be set to a specific path for the rendered output to write to. If view option is True, the plot is opened by the system default application for the image format (PDF). In IPython notebook, the returned object can be plot inlined.

Usage:

@jit
def foo():
  ...

# opens the CFG in system default application
foo.inspect_cfg(foo.signatures[0]).display(view=True)

inspect_disasm_cfg(signature=None)¶

Return a dictionary keying compiled function signatures to the control-flow graph of the disassembly of the underlying compiled ELF object. If the signature keyword is specified a control-flow graph corresponding to that individual signature is returned. This function is execution environment aware and will produce SVG output in Jupyter notebooks and ASCII in terminals.

Example:

@njit
def foo(x):
    if x < 3:
        return x + 1
    return x + 2

foo(10)

print(foo.inspect_disasm_cfg(signature=foo.signatures[0]))

Gives:

[0x08000040]>  # method.__main__.foo_241_long_long (int64_t arg1, int64_t arg3);
 ─────────────────────────────────────────────────────────────────────┐
│  0x8000040                                                          │
│ ; arg3 ; [02] -r-x section size 279 named .text                     │
│   ;-- section..text:                                                │
│   ;-- .text:                                                        │
│   ;-- __main__::foo$241(long long):                                 │
│   ;-- rip:                                                          │
│ 25: method.__main__.foo_241_long_long (int64_t arg1, int64_t arg3); │
│ ; arg int64_t arg1 @ rdi                                            │
│ ; arg int64_t arg3 @ rdx                                            │
│ ; 2                                                                 │
│ cmp rdx, 2                                                          │
│ jg 0x800004f                                                        │
└─────────────────────────────────────────────────────────────────────┘
        f t
        │ │
        │ └──────────────────────────────┐
        └──┐                             │
           │                             │
    ┌─────────────────────────┐   ┌─────────────────────────┐
    │  0x8000046              │   │  0x800004f              │
    │ ; arg3                  │   │ ; arg3                  │
    │ inc rdx                 │   │ add rdx, 2              │
    │ ; arg3                  │   │ ; arg3                  │
    │ mov qword [rdi], rdx    │   │ mov qword [rdi], rdx    │
    │ xor eax, eax            │   │ xor eax, eax            │
    │ ret                     │   │ ret                     │
    └─────────────────────────┘   └─────────────────────────┘

recompile()¶: Recompile all existing signatures. This can be useful for example if a global or closure variable was frozen by your function and its value in Python has changed. Since compiling isn’t cheap, this is mainly for testing and interactive use.

parallel_diagnostics(signature=None, level=1)¶: Print parallel diagnostic information for the given signature. If no signature is present it is printed for all known signatures. level is used to adjust the verbosity, level=1 (default) is minimum verbosity, levels 2, 3, and 4 provide increasing levels of verbosity.

get_metadata(signature=None)¶: Obtain the compilation metadata for a given signature. This is useful for developers of Numba and Numba extensions.

Vectorized functions (ufuncs and DUFuncs)¶

@numba.vectorize(*, signatures=[], identity=None, nopython=True, target='cpu', forceobj=False, cache=False, locals={})¶

Compile the decorated function and wrap it either as a Numpy ufunc or a Numba DUFunc. The optional nopython, forceobj and locals arguments have the same meaning as in numba.jit().

signatures is an optional list of signatures expressed in the same form as in the numba.jit() signature argument. If signatures is non-empty, then the decorator will compile the user Python function into a Numpy ufunc. If no signatures are given, then the decorator will wrap the user Python function in a DUFunc instance, which will compile the user function at call time whenever Numpy can not find a matching loop for the input arguments. signatures is required if target is "parallel".

identity is the identity (or unit) value of the function being implemented. Possible values are 0, 1, None, and the string "reorderable". The default is None. Both None and "reorderable" mean the function has no identity value; "reorderable" additionally specifies that reductions along multiple axes can be reordered.

If there are several signatures, they must be ordered from the more specific to the least specific. Otherwise, Numpy’s type-based dispatching may not work as expected. For example, the following is wrong:

@vectorize(["float64(float64)", "float32(float32)"])
def f(x): ...

as running it over a single-precision array will choose the float64 version of the compiled function, leading to much less efficient execution. The correct invocation is:

@vectorize(["float32(float32)", "float64(float64)"])
def f(x): ...

target is a string for backend target; Available values are “cpu”, “parallel”, and “cuda”. To use a multithreaded version, change the target to “parallel” (which requires signatures to be specified):

@vectorize(["float64(float64)", "float32(float32)"], target='parallel')
def f(x): ...

For the CUDA target, use “cuda”:

@vectorize(["float64(float64)", "float32(float32)"], target='cuda')
def f(x): ...

The compiled function can be cached to reduce future compilation time. It is enabled by setting cache to True. Only the “cpu” and “parallel” targets support caching.

@numba.guvectorize(signatures, layout, *, identity=None, nopython=True, target='cpu', forceobj=False, cache=False, locals={})¶

Generalized version of numba.vectorize(). While numba.vectorize() will produce a simple ufunc whose core functionality (the function you are decorating) operates on scalar operands and returns a scalar value, numba.guvectorize() allows you to create a Numpy ufunc whose core function takes array arguments of various dimensions.

The additional argument layout is a string specifying, in symbolic form, the dimensionality and size relationship of the argument types and return types. For example, a matrix multiplication will have a layout string of "(m,n),(n,p)->(m,p)". Its definition might be (function body omitted):

@guvectorize(["void(float64[:,:], float64[:,:], float64[:,:])"],
             "(m,n),(n,p)->(m,p)")
def f(a, b, result):
    """Fill-in *result* matrix such as result := a * b"""
    ...

If one of the arguments should be a scalar, the corresponding layout specification is () and the argument will really be given to you as a zero-dimension array (you have to dereference it to get the scalar value). For example, a one-dimension moving average with a parameterable window width may have a layout string of "(n),()->(n)".

Note that any output will be given to you preallocated as an additional function argument: your code has to fill it with the appropriate values for the function you are implementing.

If your function doesn’t take an output array, you should omit the “arrow” in the layout string (e.g. "(n),(n)"). When doing this, it is important to be aware that changes to the input arrays cannot always be relied on to be visible outside the execution of the ufunc, as NumPy may pass in temporary arrays as inputs (for example, if a cast is required).

C callbacks¶

@numba.cfunc(signature, nopython=False, cache=False, locals={})¶

Compile the decorated function on-the-fly to produce efficient machine code. The compiled code is wrapped in a thin C callback that makes it callable using the natural C ABI.

The signature is a single signature representing the signature of the C callback. It must have the same form as in jit(). The decorator does not check that the types in the signature have a well-defined representation in C.

nopython and cache are boolean flags. locals is a mapping of local variable names to Types and signatures. They all have the same meaning as in jit().

The decorator returns a CFunc object.

Note

C callbacks currently do not support object mode.

class CFunc¶

The class of objects created by cfunc(). CFunc objects expose the following attributes and methods:

address¶: The address of the compiled C callback, as an integer.

cffi¶: A cffi function pointer instance, to be passed as an argument to cffi-wrapped functions. The pointer’s type is void *, so only minimal type checking will happen when passing it to cffi.

ctypes¶: A ctypes callback instance, as if it were created using ctypes.CFUNCTYPE().

native_name¶: The name of the compiled C callback.

inspect_llvm()¶: Return the human-readable LLVM IR generated for the C callback. native_name is the name under which this callback is defined in the IR.