diff options
Diffstat (limited to 'numpy/core')
102 files changed, 4445 insertions, 2347 deletions
diff --git a/numpy/core/__init__.py b/numpy/core/__init__.py index ce443bb22..c3b3f0392 100644 --- a/numpy/core/__init__.py +++ b/numpy/core/__init__.py @@ -1,6 +1,13 @@ +""" +Contains the core of NumPy: ndarray, ufuncs, dtypes, etc. + +Please note that this module is private. All functions and objects +are available in the main ``numpy`` namespace - use that instead. + +""" + from __future__ import division, absolute_import, print_function -from .info import __doc__ from numpy.version import version as __version__ import os diff --git a/numpy/core/_add_newdocs.py b/numpy/core/_add_newdocs.py index 02700c8ca..2f1273904 100644 --- a/numpy/core/_add_newdocs.py +++ b/numpy/core/_add_newdocs.py @@ -94,7 +94,7 @@ add_newdoc('numpy.core', 'flatiter', ('coords', >>> fl = x.flat >>> fl.coords (0, 0) - >>> fl.next() + >>> next(fl) 0 >>> fl.coords (0, 1) @@ -113,7 +113,7 @@ add_newdoc('numpy.core', 'flatiter', ('index', >>> fl = x.flat >>> fl.index 0 - >>> fl.next() + >>> next(fl) 0 >>> fl.index 1 @@ -386,12 +386,12 @@ add_newdoc('numpy.core', 'nditer', >>> luf(lambda i,j:i*i + j/2, a, b) array([ 0.5, 1.5, 4.5, 9.5, 16.5]) - If operand flags `"writeonly"` or `"readwrite"` are used the operands may - be views into the original data with the `WRITEBACKIFCOPY` flag. In this case - nditer must be used as a context manager or the nditer.close - method must be called before using the result. The temporary - data will be written back to the original data when the `__exit__` - function is called but not before: + If operand flags `"writeonly"` or `"readwrite"` are used the + operands may be views into the original data with the + `WRITEBACKIFCOPY` flag. In this case `nditer` must be used as a + context manager or the `nditer.close` method must be called before + using the result. The temporary data will be written back to the + original data when the `__exit__` function is called but not before: >>> a = np.arange(6, dtype='i4')[::-2] >>> with np.nditer(a, [], @@ -413,6 +413,8 @@ add_newdoc('numpy.core', 'nditer', `x.data` will still point at some part of `a.data`, and writing to one will affect the other. + Context management and the `close` method appeared in version 1.15.0. + """) # nditer methods @@ -568,6 +570,8 @@ add_newdoc('numpy.core', 'nditer', ('close', Resolve all writeback semantics in writeable operands. + .. versionadded:: 1.15.0 + See Also -------- @@ -666,7 +670,7 @@ add_newdoc('numpy.core', 'broadcast', ('iters', >>> y = np.array([[4], [5], [6]]) >>> b = np.broadcast(x, y) >>> row, col = b.iters - >>> row.next(), col.next() + >>> next(row), next(col) (1, 4) """)) @@ -795,8 +799,7 @@ add_newdoc('numpy.core.multiarray', 'array', dtype : data-type, optional The desired data-type for the array. If not given, then the type will be determined as the minimum type required to hold the objects in the - sequence. This argument can only be used to 'upcast' the array. For - downcasting, use the .astype(t) method. + sequence. copy : bool, optional If true (default), then the object is copied. Otherwise, a copy will only be made if __array__ returns a copy, if obj is a nested sequence, @@ -1033,7 +1036,12 @@ add_newdoc('numpy.core.multiarray', 'fromstring', A string containing the data. dtype : data-type, optional The data type of the array; default: float. For binary input data, - the data must be in exactly this format. + the data must be in exactly this format. Most builtin numeric types are + supported and extension types may be supported. + + .. versionadded:: 1.18.0 + Complex dtypes. + count : int, optional Read this number of `dtype` elements from the data. If this is negative (the default), the count will be determined from the @@ -1169,6 +1177,11 @@ add_newdoc('numpy.core.multiarray', 'fromfile', Data type of the returned array. For binary files, it is used to determine the size and byte-order of the items in the file. + Most builtin numeric types are supported and extension types may be supported. + + .. versionadded:: 1.18.0 + Complex dtypes. + count : int Number of items to read. ``-1`` means all items (i.e., the complete file). @@ -1193,7 +1206,7 @@ add_newdoc('numpy.core.multiarray', 'fromfile', Notes ----- Do not rely on the combination of `tofile` and `fromfile` for - data storage, as the binary files generated are are not platform + data storage, as the binary files generated are not platform independent. In particular, no byte-order or data-type information is saved. Data can be stored in the platform independent ``.npy`` format using `save` and `load` instead. @@ -1323,9 +1336,9 @@ add_newdoc('numpy.core.multiarray', 'arange', See Also -------- - linspace : Evenly spaced numbers with careful handling of endpoints. - ogrid: Arrays of evenly spaced numbers in N-dimensions. - mgrid: Grid-shaped arrays of evenly spaced numbers in N-dimensions. + numpy.linspace : Evenly spaced numbers with careful handling of endpoints. + numpy.ogrid: Arrays of evenly spaced numbers in N-dimensions. + numpy.mgrid: Grid-shaped arrays of evenly spaced numbers in N-dimensions. Examples -------- @@ -1343,7 +1356,7 @@ add_newdoc('numpy.core.multiarray', 'arange', add_newdoc('numpy.core.multiarray', '_get_ndarray_c_version', """_get_ndarray_c_version() - Return the compile time NDARRAY_VERSION number. + Return the compile time NPY_VERSION (formerly called NDARRAY_VERSION) number. """) @@ -2736,6 +2749,8 @@ add_newdoc('numpy.core.multiarray', 'ndarray', ('byteswap', Toggle between low-endian and big-endian data representation by returning a byteswapped array, optionally swapped in-place. + Arrays of byte-strings are not swapped. The real and imaginary + parts of a complex number are swapped individually. Parameters ---------- @@ -2758,13 +2773,24 @@ add_newdoc('numpy.core.multiarray', 'ndarray', ('byteswap', >>> list(map(hex, A)) ['0x100', '0x1', '0x3322'] - Arrays of strings are not swapped + Arrays of byte-strings are not swapped - >>> A = np.array(['ceg', 'fac']) + >>> A = np.array([b'ceg', b'fac']) >>> A.byteswap() - Traceback (most recent call last): - ... - UnicodeDecodeError: ... + array([b'ceg', b'fac'], dtype='|S3') + + ``A.newbyteorder().byteswap()`` produces an array with the same values + but different representation in memory + + >>> A = np.array([1, 2, 3]) + >>> A.view(np.uint8) + array([1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, + 0, 0], dtype=uint8) + >>> A.newbyteorder().byteswap(inplace=True) + array([1, 2, 3]) + >>> A.view(np.uint8) + array([0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, + 0, 3], dtype=uint8) """)) @@ -3690,10 +3716,10 @@ add_newdoc('numpy.core.multiarray', 'ndarray', ('sort', See Also -------- numpy.sort : Return a sorted copy of an array. - argsort : Indirect sort. - lexsort : Indirect stable sort on multiple keys. - searchsorted : Find elements in sorted array. - partition: Partial sort. + numpy.argsort : Indirect sort. + numpy.lexsort : Indirect stable sort on multiple keys. + numpy.searchsorted : Find elements in sorted array. + numpy.partition: Partial sort. Notes ----- @@ -3927,15 +3953,22 @@ add_newdoc('numpy.core.multiarray', 'ndarray', ('tolist', Examples -------- - For a 1D array, ``a.tolist()`` is almost the same as ``list(a)``: + For a 1D array, ``a.tolist()`` is almost the same as ``list(a)``, + except that ``tolist`` changes numpy scalars to Python scalars: - >>> a = np.array([1, 2]) - >>> list(a) + >>> a = np.uint32([1, 2]) + >>> a_list = list(a) + >>> a_list [1, 2] - >>> a.tolist() + >>> type(a_list[0]) + <class 'numpy.uint32'> + >>> a_tolist = a.tolist() + >>> a_tolist [1, 2] + >>> type(a_tolist[0]) + <class 'int'> - However, for a 2D array, ``tolist`` applies recursively: + Additionally, for a 2D array, ``tolist`` applies recursively: >>> a = np.array([[1, 2], [3, 4]]) >>> list(a) @@ -4220,7 +4253,7 @@ add_newdoc('numpy.core.umath', 'frompyfunc', See Also -------- - vectorize : evaluates pyfunc over input arrays using broadcasting rules of numpy + vectorize : Evaluates pyfunc over input arrays using broadcasting rules of numpy. Notes ----- @@ -4481,7 +4514,7 @@ add_newdoc('numpy.core', 'ufunc', Alternate array object(s) in which to put the result; if provided, it must have a shape that the inputs broadcast to. A tuple of arrays (possible only as a keyword argument) must have length equal to the - number of outputs; use `None` for uninitialized outputs to be + number of outputs; use None for uninitialized outputs to be allocated by the ufunc. where : array_like, optional This condition is broadcast over the input. At locations where the @@ -4675,7 +4708,7 @@ add_newdoc('numpy.core', 'ufunc', ('signature', ----- Generalized ufuncs are used internally in many linalg functions, and in the testing suite; the examples below are taken from these. - For ufuncs that operate on scalars, the signature is `None`, which is + For ufuncs that operate on scalars, the signature is None, which is equivalent to '()' for every argument. Examples @@ -4726,7 +4759,7 @@ add_newdoc('numpy.core', 'ufunc', ('reduce', .. versionadded:: 1.7.0 - If this is `None`, a reduction is performed over all the axes. + If this is None, a reduction is performed over all the axes. If this is a tuple of ints, a reduction is performed on multiple axes, instead of a single axis or all the axes as before. @@ -4739,7 +4772,7 @@ add_newdoc('numpy.core', 'ufunc', ('reduce', to the data-type of the output array if this is provided, or the data-type of the input array if no output array is provided. out : ndarray, None, or tuple of ndarray and None, optional - A location into which the result is stored. If not provided or `None`, + A location into which the result is stored. If not provided or None, a freshly-allocated array is returned. For consistency with ``ufunc.__call__``, if given as a keyword, this may be wrapped in a 1-element tuple. @@ -4856,7 +4889,7 @@ add_newdoc('numpy.core', 'ufunc', ('accumulate', to the data-type of the output array if such is provided, or the the data-type of the input array if no output array is provided. out : ndarray, None, or tuple of ndarray and None, optional - A location into which the result is stored. If not provided or `None`, + A location into which the result is stored. If not provided or None, a freshly-allocated array is returned. For consistency with ``ufunc.__call__``, if given as a keyword, this may be wrapped in a 1-element tuple. @@ -4938,7 +4971,7 @@ add_newdoc('numpy.core', 'ufunc', ('reduceat', to the data type of the output array if this is provided, or the data type of the input array if no output array is provided. out : ndarray, None, or tuple of ndarray and None, optional - A location into which the result is stored. If not provided or `None`, + A location into which the result is stored. If not provided or None, a freshly-allocated array is returned. For consistency with ``ufunc.__call__``, if given as a keyword, this may be wrapped in a 1-element tuple. @@ -5311,7 +5344,8 @@ add_newdoc('numpy.core.multiarray', 'dtype', ('descr', `__array_interface__` attribute. Warning: This attribute exists specifically for `__array_interface__`, - and is not a datatype description compatible with `np.dtype`. + and passing it directly to `np.dtype` will not accurately reconstruct + some dtypes (e.g., scalar and subarray dtypes). Examples -------- @@ -6837,3 +6871,4 @@ for float_name in ('half', 'single', 'double', 'longdouble'): >>> np.{ftype}(-.25).as_integer_ratio() (-1, 4) """.format(ftype=float_name))) + diff --git a/numpy/core/_aliased_types.py b/numpy/core/_aliased_types.py deleted file mode 100644 index e69de29bb..000000000 --- a/numpy/core/_aliased_types.py +++ /dev/null diff --git a/numpy/core/_dtype.py b/numpy/core/_dtype.py index 3a12c8fad..df1ff180e 100644 --- a/numpy/core/_dtype.py +++ b/numpy/core/_dtype.py @@ -252,7 +252,7 @@ def _is_packed(dtype): from a list of the field names and dtypes with no additional dtype parameters. - Duplicates the C `is_dtype_struct_simple_unaligned_layout` functio. + Duplicates the C `is_dtype_struct_simple_unaligned_layout` function. """ total_offset = 0 for name in dtype.names: @@ -316,26 +316,39 @@ def _subarray_str(dtype): ) +def _name_includes_bit_suffix(dtype): + if dtype.type == np.object_: + # pointer size varies by system, best to omit it + return False + elif dtype.type == np.bool_: + # implied + return False + elif np.issubdtype(dtype, np.flexible) and _isunsized(dtype): + # unspecified + return False + else: + return True + + def _name_get(dtype): - # provides dtype.name.__get__ + # provides dtype.name.__get__, documented as returning a "bit name" if dtype.isbuiltin == 2: # user dtypes don't promise to do anything special return dtype.type.__name__ - # Builtin classes are documented as returning a "bit name" - name = dtype.type.__name__ - - # handle bool_, str_, etc - if name[-1] == '_': - name = name[:-1] + if issubclass(dtype.type, np.void): + # historically, void subclasses preserve their name, eg `record64` + name = dtype.type.__name__ + else: + name = _kind_name(dtype) - # append bit counts to str, unicode, and void - if np.issubdtype(dtype, np.flexible) and not _isunsized(dtype): + # append bit counts + if _name_includes_bit_suffix(dtype): name += "{}".format(dtype.itemsize * 8) # append metadata to datetimes - elif dtype.type in (np.datetime64, np.timedelta64): + if dtype.type in (np.datetime64, np.timedelta64): name += _datetime_metadata_str(dtype) return name diff --git a/numpy/core/_exceptions.py b/numpy/core/_exceptions.py index b3805af04..88a45561f 100644 --- a/numpy/core/_exceptions.py +++ b/numpy/core/_exceptions.py @@ -147,6 +147,54 @@ class _ArrayMemoryError(MemoryError): self.shape = shape self.dtype = dtype - def __str__(self): - return "Unable to allocate array with shape {} and data type {}".format(self.shape, self.dtype) + @property + def _total_size(self): + num_bytes = self.dtype.itemsize + for dim in self.shape: + num_bytes *= dim + return num_bytes + + @staticmethod + def _size_to_string(num_bytes): + """ Convert a number of bytes into a binary size string """ + import math + + # https://en.wikipedia.org/wiki/Binary_prefix + LOG2_STEP = 10 + STEP = 1024 + units = ['bytes', 'KiB', 'MiB', 'GiB', 'TiB', 'PiB', 'EiB'] + + unit_i = max(num_bytes.bit_length() - 1, 1) // LOG2_STEP + unit_val = 1 << (unit_i * LOG2_STEP) + n_units = num_bytes / unit_val + del unit_val + + # ensure we pick a unit that is correct after rounding + if round(n_units) == STEP: + unit_i += 1 + n_units /= STEP + + # deal with sizes so large that we don't have units for them + if unit_i >= len(units): + new_unit_i = len(units) - 1 + n_units *= 1 << ((unit_i - new_unit_i) * LOG2_STEP) + unit_i = new_unit_i + + unit_name = units[unit_i] + # format with a sensible number of digits + if unit_i == 0: + # no decimal point on bytes + return '{:.0f} {}'.format(n_units, unit_name) + elif round(n_units) < 1000: + # 3 significant figures, if none are dropped to the left of the . + return '{:#.3g} {}'.format(n_units, unit_name) + else: + # just give all the digits otherwise + return '{:#.0f} {}'.format(n_units, unit_name) + def __str__(self): + size_str = self._size_to_string(self._total_size) + return ( + "Unable to allocate {} for an array with shape {} and data type {}" + .format(size_str, self.shape, self.dtype) + ) diff --git a/numpy/core/_internal.py b/numpy/core/_internal.py index 5fd643505..05e401e0b 100644 --- a/numpy/core/_internal.py +++ b/numpy/core/_internal.py @@ -313,7 +313,7 @@ class _ctypes(object): crashing. User Beware! The value of this attribute is exactly the same as ``self._array_interface_['data'][0]``. - Note that unlike `data_as`, a reference will not be kept to the array: + Note that unlike ``data_as``, a reference will not be kept to the array: code like ``ctypes.c_void_p((a + b).ctypes.data)`` will result in a pointer to a deallocated array, and should be spelt ``(a + b).ctypes.data_as(ctypes.c_void_p)`` diff --git a/numpy/core/arrayprint.py b/numpy/core/arrayprint.py index 8251beae8..401018015 100644 --- a/numpy/core/arrayprint.py +++ b/numpy/core/arrayprint.py @@ -89,8 +89,10 @@ def _make_options_dict(precision=None, threshold=None, edgeitems=None, "`False`", stacklevel=3) if threshold is not None: # forbid the bad threshold arg suggested by stack overflow, gh-12351 - if not isinstance(threshold, numbers.Number) or np.isnan(threshold): - raise ValueError("threshold must be numeric and non-NAN, try " + if not isinstance(threshold, numbers.Number): + raise TypeError("threshold must be numeric") + if np.isnan(threshold): + raise ValueError("threshold must be non-NAN, try " "sys.maxsize for untruncated representation") return options @@ -109,11 +111,12 @@ def set_printoptions(precision=None, threshold=None, edgeitems=None, ---------- precision : int or None, optional Number of digits of precision for floating point output (default 8). - May be `None` if `floatmode` is not `fixed`, to print as many digits as + May be None if `floatmode` is not `fixed`, to print as many digits as necessary to uniquely specify the value. threshold : int, optional Total number of array elements which trigger summarization rather than full repr (default 1000). + To always use the full repr without summarization, pass `sys.maxsize`. edgeitems : int, optional Number of array items in summary at beginning and end of each dimension (default 3). @@ -191,12 +194,14 @@ def set_printoptions(precision=None, threshold=None, edgeitems=None, See Also -------- - get_printoptions, set_string_function, array2string + get_printoptions, printoptions, set_string_function, array2string Notes ----- `formatter` is always reset with a call to `set_printoptions`. + Use `printoptions` as a context manager to set the values temporarily. + Examples -------- Floating point precision can be set: @@ -233,9 +238,16 @@ def set_printoptions(precision=None, threshold=None, edgeitems=None, To put back the default options, you can use: - >>> np.set_printoptions(edgeitems=3,infstr='inf', + >>> np.set_printoptions(edgeitems=3, infstr='inf', ... linewidth=75, nanstr='nan', precision=8, ... suppress=False, threshold=1000, formatter=None) + + Also to temporarily override options, use `printoptions` as a context manager: + + >>> with np.printoptions(precision=2, suppress=True, threshold=5): + ... np.linspace(0, 10, 10) + array([ 0. , 1.11, 2.22, ..., 7.78, 8.89, 10. ]) + """ legacy = kwarg.pop('legacy', None) if kwarg: @@ -282,7 +294,7 @@ def get_printoptions(): See Also -------- - set_printoptions, set_string_function + set_printoptions, printoptions, set_string_function """ return _format_options.copy() @@ -981,20 +993,6 @@ class FloatingFormat(object): pad_left=self.pad_left, pad_right=self.pad_right) -# for back-compatibility, we keep the classes for each float type too -class FloatFormat(FloatingFormat): - def __init__(self, *args, **kwargs): - warnings.warn("FloatFormat has been replaced by FloatingFormat", - DeprecationWarning, stacklevel=2) - super(FloatFormat, self).__init__(*args, **kwargs) - - -class LongFloatFormat(FloatingFormat): - def __init__(self, *args, **kwargs): - warnings.warn("LongFloatFormat has been replaced by FloatingFormat", - DeprecationWarning, stacklevel=2) - super(LongFloatFormat, self).__init__(*args, **kwargs) - @set_module('numpy') def format_float_scientific(x, precision=None, unique=True, trim='k', @@ -1193,21 +1191,6 @@ class ComplexFloatingFormat(object): return r + i -# for back-compatibility, we keep the classes for each complex type too -class ComplexFormat(ComplexFloatingFormat): - def __init__(self, *args, **kwargs): - warnings.warn( - "ComplexFormat has been replaced by ComplexFloatingFormat", - DeprecationWarning, stacklevel=2) - super(ComplexFormat, self).__init__(*args, **kwargs) - -class LongComplexFormat(ComplexFloatingFormat): - def __init__(self, *args, **kwargs): - warnings.warn( - "LongComplexFormat has been replaced by ComplexFloatingFormat", - DeprecationWarning, stacklevel=2) - super(LongComplexFormat, self).__init__(*args, **kwargs) - class _TimelikeFormat(object): def __init__(self, data): @@ -1318,16 +1301,6 @@ class StructuredVoidFormat(object): return "({})".format(", ".join(str_fields)) -# for backwards compatibility -class StructureFormat(StructuredVoidFormat): - def __init__(self, *args, **kwargs): - # NumPy 1.14, 2018-02-14 - warnings.warn( - "StructureFormat has been replaced by StructuredVoidFormat", - DeprecationWarning, stacklevel=2) - super(StructureFormat, self).__init__(*args, **kwargs) - - def _void_scalar_repr(x): """ Implements the repr for structured-void scalars. It is called from the @@ -1506,7 +1479,11 @@ def array_repr(arr, max_line_width=None, precision=None, suppress_small=None): arr, max_line_width, precision, suppress_small) -_guarded_str = _recursive_guard()(str) +@_recursive_guard() +def _guarded_repr_or_str(v): + if isinstance(v, bytes): + return repr(v) + return str(v) def _array_str_implementation( @@ -1524,7 +1501,7 @@ def _array_str_implementation( # obtain a scalar and call str on it, avoiding problems for subclasses # for which indexing with () returns a 0d instead of a scalar by using # ndarray's getindex. Also guard against recursive 0d object arrays. - return _guarded_str(np.ndarray.__getitem__(a, ())) + return _guarded_repr_or_str(np.ndarray.__getitem__(a, ())) return array2string(a, max_line_width, precision, suppress_small, ' ', "") @@ -1641,5 +1618,5 @@ def set_string_function(f, repr=True): else: return multiarray.set_string_function(f, repr) -set_string_function(_default_array_str, 0) -set_string_function(_default_array_repr, 1) +set_string_function(_default_array_str, False) +set_string_function(_default_array_repr, True) diff --git a/numpy/core/code_generators/cversions.txt b/numpy/core/code_generators/cversions.txt index 00f10df57..72d2af8b9 100644 --- a/numpy/core/code_generators/cversions.txt +++ b/numpy/core/code_generators/cversions.txt @@ -47,4 +47,6 @@ # Deprecate PyArray_SetNumericOps and PyArray_GetNumericOps, # Add fields core_dim_flags and core_dim_sizes to PyUFuncObject. # Add PyUFunc_FromFuncAndDataAndSignatureAndIdentity to ufunc_funcs_api. +# Version 13 (NumPy 1.17) No change. +# Version 13 (NumPy 1.18) No change. 0x0000000d = 5b0e8bbded00b166125974fc71e80a33 diff --git a/numpy/core/code_generators/genapi.py b/numpy/core/code_generators/genapi.py index 923c34425..22afa0320 100644 --- a/numpy/core/code_generators/genapi.py +++ b/numpy/core/code_generators/genapi.py @@ -8,8 +8,11 @@ specified. """ from __future__ import division, absolute_import, print_function +from numpy.distutils.conv_template import process_file as process_c_file + import sys, os, re import hashlib +import io import textwrap @@ -215,7 +218,10 @@ def find_functions(filename, tag='API'): This function does foo... */ """ - fo = open(filename, 'r') + if filename.endswith(('.c.src', '.h.src')): + fo = io.StringIO(process_c_file(filename)) + else: + fo = open(filename, 'r') functions = [] return_type = None function_name = None @@ -259,7 +265,8 @@ def find_functions(filename, tag='API'): elif state == STATE_ARGS: if line.startswith('{'): # finished - fargs_str = ' '.join(function_args).rstrip(' )') + # remove any white space and the closing bracket: + fargs_str = ' '.join(function_args).rstrip()[:-1].rstrip() fargs = split_arguments(fargs_str) f = Function(function_name, return_type, fargs, '\n'.join(doclist)) diff --git a/numpy/core/code_generators/generate_umath.py b/numpy/core/code_generators/generate_umath.py index bf1747272..6d76f7ca2 100644 --- a/numpy/core/code_generators/generate_umath.py +++ b/numpy/core/code_generators/generate_umath.py @@ -226,7 +226,9 @@ chartoname = { 'P': 'OBJECT', } -all = '?bBhHiIlLqQefdgFDGOMm' +noobj = '?bBhHiIlLqQefdgFDGmM' +all = '?bBhHiIlLqQefdgFDGOmM' + O = 'O' P = 'P' ints = 'bBhHiIlLqQ' @@ -246,10 +248,8 @@ inexactvec = 'fd' noint = inexact+O nointP = inexact+P allP = bints+times+flts+cmplxP -nobool = all[1:] -noobj = all[:-3]+all[-2:] -nobool_or_obj = all[1:-3]+all[-2:] -nobool_or_datetime = all[1:-2]+all[-1:] +nobool_or_obj = noobj[1:] +nobool_or_datetime = noobj[1:-1] + O # includes m - timedelta64 intflt = ints+flts intfltcmplx = ints+flts+cmplx nocmplx = bints+times+flts @@ -287,7 +287,7 @@ defdict = { Ufunc(2, 1, None, # Zero is only a unit to the right, not the left docstrings.get('numpy.core.umath.subtract'), 'PyUFunc_SubtractionTypeResolver', - TD(notimes_or_obj, simd=[('avx2', ints)]), + TD(ints + inexact, simd=[('avx2', ints)]), [TypeDescription('M', FullTypeDescr, 'Mm', 'M'), TypeDescription('m', FullTypeDescr, 'mm', 'm'), TypeDescription('M', FullTypeDescr, 'MM', 'm'), @@ -358,14 +358,14 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.square'), None, - TD(ints+inexact, simd=[('avx2', ints)]), + TD(ints+inexact, simd=[('avx2', ints), ('fma', 'fd'), ('avx512f', 'fd')]), TD(O, f='Py_square'), ), 'reciprocal': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.reciprocal'), None, - TD(ints+inexact, simd=[('avx2', ints)]), + TD(ints+inexact, simd=[('avx2', ints), ('fma', 'fd'), ('avx512f','fd')]), TD(O, f='Py_reciprocal'), ), # This is no longer used as numpy.ones_like, however it is @@ -395,7 +395,7 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.absolute'), 'PyUFunc_AbsoluteTypeResolver', - TD(bints+flts+timedeltaonly), + TD(bints+flts+timedeltaonly, simd=[('fma', 'fd'), ('avx512f', 'fd')]), TD(cmplx, out=('f', 'd', 'g')), TD(O, f='PyNumber_Absolute'), ), @@ -409,7 +409,7 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.negative'), 'PyUFunc_NegativeTypeResolver', - TD(bints+flts+timedeltaonly, simd=[('avx2', ints)]), + TD(ints+flts+timedeltaonly, simd=[('avx2', ints)]), TD(cmplx, f='neg'), TD(O, f='PyNumber_Negative'), ), @@ -433,6 +433,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'greater_equal': Ufunc(2, 1, None, @@ -440,6 +441,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'less': Ufunc(2, 1, None, @@ -447,6 +449,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'less_equal': Ufunc(2, 1, None, @@ -454,6 +457,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'equal': Ufunc(2, 1, None, @@ -461,6 +465,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'not_equal': Ufunc(2, 1, None, @@ -468,6 +473,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(all, out='?', simd=[('avx2', ints)]), [TypeDescription('O', FullTypeDescr, 'OO', 'O')], + TD('O', out='?'), ), 'logical_and': Ufunc(2, 1, True_, @@ -475,6 +481,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(nodatetime_or_obj, out='?', simd=[('avx2', ints)]), TD(O, f='npy_ObjectLogicalAnd'), + TD(O, f='npy_ObjectLogicalAnd', out='?'), ), 'logical_not': Ufunc(1, 1, None, @@ -482,6 +489,7 @@ defdict = { None, TD(nodatetime_or_obj, out='?', simd=[('avx2', ints)]), TD(O, f='npy_ObjectLogicalNot'), + TD(O, f='npy_ObjectLogicalNot', out='?'), ), 'logical_or': Ufunc(2, 1, False_, @@ -489,6 +497,7 @@ defdict = { 'PyUFunc_SimpleBinaryComparisonTypeResolver', TD(nodatetime_or_obj, out='?', simd=[('avx2', ints)]), TD(O, f='npy_ObjectLogicalOr'), + TD(O, f='npy_ObjectLogicalOr', out='?'), ), 'logical_xor': Ufunc(2, 1, False_, @@ -662,14 +671,18 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.cos'), None, - TD(inexact, f='cos', astype={'e':'f'}), + TD('e', f='cos', astype={'e':'f'}), + TD('f', simd=[('fma', 'f'), ('avx512f', 'f')]), + TD('fdg' + cmplx, f='cos'), TD(P, f='cos'), ), 'sin': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.sin'), None, - TD(inexact, f='sin', astype={'e':'f'}), + TD('e', f='sin', astype={'e':'f'}), + TD('f', simd=[('fma', 'f'), ('avx512f', 'f')]), + TD('fdg' + cmplx, f='sin'), TD(P, f='sin'), ), 'tan': @@ -705,8 +718,8 @@ defdict = { docstrings.get('numpy.core.umath.exp'), None, TD('e', f='exp', astype={'e':'f'}), - TD('f', simd=[('avx2', 'f'), ('avx512f', 'f')]), - TD(inexact, f='exp', astype={'e':'f'}), + TD('f', simd=[('fma', 'f'), ('avx512f', 'f')]), + TD('fdg' + cmplx, f='exp'), TD(P, f='exp'), ), 'exp2': @@ -728,8 +741,8 @@ defdict = { docstrings.get('numpy.core.umath.log'), None, TD('e', f='log', astype={'e':'f'}), - TD('f', simd=[('avx2', 'f'), ('avx512f', 'f')]), - TD(inexact, f='log', astype={'e':'f'}), + TD('f', simd=[('fma', 'f'), ('avx512f', 'f')]), + TD('fdg' + cmplx, f='log'), TD(P, f='log'), ), 'log2': @@ -758,8 +771,8 @@ defdict = { docstrings.get('numpy.core.umath.sqrt'), None, TD('e', f='sqrt', astype={'e':'f'}), - TD(inexactvec), - TD(inexact, f='sqrt', astype={'e':'f'}), + TD(inexactvec, simd=[('fma', 'fd'), ('avx512f', 'fd')]), + TD('fdg' + cmplx, f='sqrt'), TD(P, f='sqrt'), ), 'cbrt': @@ -773,14 +786,18 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.ceil'), None, - TD(flts, f='ceil', astype={'e':'f'}), + TD('e', f='ceil', astype={'e':'f'}), + TD(inexactvec, simd=[('fma', 'fd'), ('avx512f', 'fd')]), + TD('fdg', f='ceil'), TD(O, f='npy_ObjectCeil'), ), 'trunc': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.trunc'), None, - TD(flts, f='trunc', astype={'e':'f'}), + TD('e', f='trunc', astype={'e':'f'}), + TD(inexactvec, simd=[('fma', 'fd'), ('avx512f', 'fd')]), + TD('fdg', f='trunc'), TD(O, f='npy_ObjectTrunc'), ), 'fabs': @@ -794,14 +811,18 @@ defdict = { Ufunc(1, 1, None, docstrings.get('numpy.core.umath.floor'), None, - TD(flts, f='floor', astype={'e':'f'}), + TD('e', f='floor', astype={'e':'f'}), + TD(inexactvec, simd=[('fma', 'fd'), ('avx512f', 'fd')]), + TD('fdg', f='floor'), TD(O, f='npy_ObjectFloor'), ), 'rint': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.rint'), None, - TD(inexact, f='rint', astype={'e':'f'}), + TD('e', f='rint', astype={'e':'f'}), + TD(inexactvec, simd=[('fma', 'fd'), ('avx512f', 'fd')]), + TD('fdg' + cmplx, f='rint'), TD(P, f='rint'), ), 'arctan2': @@ -837,8 +858,8 @@ defdict = { 'isnan': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.isnan'), - None, - TD(nodatetime_or_obj, out='?'), + 'PyUFunc_IsFiniteTypeResolver', + TD(noobj, out='?'), ), 'isnat': Ufunc(1, 1, None, @@ -849,8 +870,8 @@ defdict = { 'isinf': Ufunc(1, 1, None, docstrings.get('numpy.core.umath.isinf'), - None, - TD(nodatetime_or_obj, out='?'), + 'PyUFunc_IsFiniteTypeResolver', + TD(noobj, out='?'), ), 'isfinite': Ufunc(1, 1, None, diff --git a/numpy/core/code_generators/ufunc_docstrings.py b/numpy/core/code_generators/ufunc_docstrings.py index fb418aadc..4dec73505 100644 --- a/numpy/core/code_generators/ufunc_docstrings.py +++ b/numpy/core/code_generators/ufunc_docstrings.py @@ -22,7 +22,7 @@ subst = { 'PARAMS': textwrap.dedent(""" out : ndarray, None, or tuple of ndarray and None, optional A location into which the result is stored. If provided, it must have - a shape that the inputs broadcast to. If not provided or `None`, + a shape that the inputs broadcast to. If not provided or None, a freshly-allocated array is returned. A tuple (possible only as a keyword argument) must have length equal to the number of outputs. where : array_like, optional @@ -2183,7 +2183,7 @@ add_newdoc('numpy.core.umath', 'logical_and', Returns ------- y : ndarray or bool - Boolean result of the logical OR operation applied to the elements + Boolean result of the logical AND operation applied to the elements of `x1` and `x2`; the shape is determined by broadcasting. $OUT_SCALAR_2 @@ -2596,7 +2596,7 @@ add_newdoc('numpy.core.umath', 'matmul', out : ndarray, optional A location into which the result is stored. If provided, it must have a shape that matches the signature `(n,k),(k,m)->(n,m)`. If not - provided or `None`, a freshly-allocated array is returned. + provided or None, a freshly-allocated array is returned. **kwargs For other keyword-only arguments, see the :ref:`ufunc docs <ufuncs.kwargs>`. diff --git a/numpy/core/defchararray.py b/numpy/core/defchararray.py index d7ecce1b4..2d89d6fe0 100644 --- a/numpy/core/defchararray.py +++ b/numpy/core/defchararray.py @@ -82,7 +82,7 @@ def _clean_args(*args): Many of the Python string operations that have optional arguments do not use 'None' to indicate a default value. In these cases, - we need to remove all `None` arguments, and those following them. + we need to remove all None arguments, and those following them. """ newargs = [] for chk in args: @@ -1333,7 +1333,7 @@ def rsplit(a, sep=None, maxsplit=None): a : array_like of str or unicode sep : str or unicode, optional - If `sep` is not specified or `None`, any whitespace string + If `sep` is not specified or None, any whitespace string is a separator. maxsplit : int, optional If `maxsplit` is given, at most `maxsplit` splits are done, @@ -1417,7 +1417,7 @@ def split(a, sep=None, maxsplit=None): a : array_like of str or unicode sep : str or unicode, optional - If `sep` is not specified or `None`, any whitespace string is a + If `sep` is not specified or None, any whitespace string is a separator. maxsplit : int, optional @@ -1840,7 +1840,7 @@ class chararray(ndarray): This constructor creates the array, using `buffer` (with `offset` and `strides`) if it is not ``None``. If `buffer` is ``None``, then constructs a new array with `strides` in "C order", unless both - ``len(shape) >= 2`` and ``order='Fortran'``, in which case `strides` + ``len(shape) >= 2`` and ``order='F'``, in which case `strides` is in "Fortran order". Methods @@ -2659,7 +2659,7 @@ def array(obj, itemsize=None, copy=True, unicode=None, order=None): unicode : bool, optional When true, the resulting `chararray` can contain Unicode characters, when false only 8-bit characters. If unicode is - `None` and `obj` is one of the following: + None and `obj` is one of the following: - a `chararray`, - an ndarray of type `str` or `unicode` @@ -2799,7 +2799,7 @@ def asarray(obj, itemsize=None, unicode=None, order=None): unicode : bool, optional When true, the resulting `chararray` can contain Unicode characters, when false only 8-bit characters. If unicode is - `None` and `obj` is one of the following: + None and `obj` is one of the following: - a `chararray`, - an ndarray of type `str` or 'unicode` diff --git a/numpy/core/fromnumeric.py b/numpy/core/fromnumeric.py index 08f17aae4..d454480a8 100644 --- a/numpy/core/fromnumeric.py +++ b/numpy/core/fromnumeric.py @@ -25,7 +25,7 @@ __all__ = [ 'argmin', 'argpartition', 'argsort', 'around', 'choose', 'clip', 'compress', 'cumprod', 'cumproduct', 'cumsum', 'diagonal', 'mean', 'ndim', 'nonzero', 'partition', 'prod', 'product', 'ptp', 'put', - 'rank', 'ravel', 'repeat', 'reshape', 'resize', 'round_', + 'ravel', 'repeat', 'reshape', 'resize', 'round_', 'searchsorted', 'shape', 'size', 'sometrue', 'sort', 'squeeze', 'std', 'sum', 'swapaxes', 'take', 'trace', 'transpose', 'var', ] @@ -380,6 +380,7 @@ def choose(a, choices, out=None, mode='raise'): See Also -------- ndarray.choose : equivalent method + numpy.take_along_axis : Preferable if `choices` is an array Notes ----- @@ -795,7 +796,9 @@ def argpartition(a, kth, axis=-1, kind='introselect', order=None): -------- partition : Describes partition algorithms used. ndarray.partition : Inplace partition. - argsort : Full indirect sort + argsort : Full indirect sort. + take_along_axis : Apply ``index_array`` from argpartition + to an array as if by calling partition. Notes ----- @@ -815,6 +818,14 @@ def argpartition(a, kth, axis=-1, kind='introselect', order=None): >>> np.array(x)[np.argpartition(x, 3)] array([2, 1, 3, 4]) + Multi-dimensional array: + + >>> x = np.array([[3, 4, 2], [1, 3, 1]]) + >>> index_array = np.argpartition(x, kth=1, axis=-1) + >>> np.take_along_axis(x, index_array, axis=-1) # same as np.partition(x, kth=1) + array([[2, 3, 4], + [1, 1, 3]]) + """ return _wrapfunc(a, 'argpartition', kth, axis=axis, kind=kind, order=order) @@ -908,14 +919,18 @@ def sort(a, axis=-1, kind=None, order=None): .. versionadded:: 1.12.0 - quicksort has been changed to an introsort which will switch - heapsort when it does not make enough progress. This makes its - worst case O(n*log(n)). - - 'stable' automatically choses the best stable sorting algorithm - for the data type being sorted. It, along with 'mergesort' is - currently mapped to timsort or radix sort depending on the - data type. API forward compatibility currently limits the + quicksort has been changed to `introsort <https://en.wikipedia.org/wiki/Introsort>`_. + When sorting does not make enough progress it switches to + `heapsort <https://en.wikipedia.org/wiki/Heapsort>`_. + This implementation makes quicksort O(n*log(n)) in the worst case. + + 'stable' automatically chooses the best stable sorting algorithm + for the data type being sorted. + It, along with 'mergesort' is currently mapped to + `timsort <https://en.wikipedia.org/wiki/Timsort>`_ + or `radix sort <https://en.wikipedia.org/wiki/Radix_sort>`_ + depending on the data type. + API forward compatibility currently limits the ability to select the implementation and it is hardwired for the different data types. @@ -924,11 +939,15 @@ def sort(a, axis=-1, kind=None, order=None): Timsort is added for better performance on already or nearly sorted data. On random data timsort is almost identical to mergesort. It is now used for stable sort while quicksort is still the - default sort if none is chosen. For details of timsort, refer to + default sort if none is chosen. For timsort details, refer to `CPython listsort.txt <https://github.com/python/cpython/blob/3.7/Objects/listsort.txt>`_. 'mergesort' and 'stable' are mapped to radix sort for integer data types. Radix sort is an O(n) sort instead of O(n log n). + .. versionchanged:: 1.17.0 + + NaT now sorts to the end of arrays for consistency with NaN. + Examples -------- >>> a = np.array([[1,4],[3,1]]) @@ -1020,6 +1039,8 @@ def argsort(a, axis=-1, kind=None, order=None): lexsort : Indirect stable sort with multiple keys. ndarray.sort : Inplace sort. argpartition : Indirect partial sort. + take_along_axis : Apply ``index_array`` from argsort + to an array as if by calling sort. Notes ----- @@ -1115,6 +1136,8 @@ def argmax(a, axis=None, out=None): ndarray.argmax, argmin amax : The maximum value along a given axis. unravel_index : Convert a flat index into an index tuple. + take_along_axis : Apply ``np.expand_dims(index_array, axis)`` + from argmax to an array as if by calling max. Notes ----- @@ -1149,6 +1172,16 @@ def argmax(a, axis=None, out=None): >>> np.argmax(b) # Only the first occurrence is returned. 1 + >>> x = np.array([[4,2,3], [1,0,3]]) + >>> index_array = np.argmax(x, axis=-1) + >>> # Same as np.max(x, axis=-1, keepdims=True) + >>> np.take_along_axis(x, np.expand_dims(index_array, axis=-1), axis=-1) + array([[4], + [3]]) + >>> # Same as np.max(x, axis=-1) + >>> np.take_along_axis(x, np.expand_dims(index_array, axis=-1), axis=-1).squeeze(axis=-1) + array([4, 3]) + """ return _wrapfunc(a, 'argmax', axis=axis, out=out) @@ -1184,6 +1217,8 @@ def argmin(a, axis=None, out=None): ndarray.argmin, argmax amin : The minimum value along a given axis. unravel_index : Convert a flat index into an index tuple. + take_along_axis : Apply ``np.expand_dims(index_array, axis)`` + from argmin to an array as if by calling min. Notes ----- @@ -1218,6 +1253,16 @@ def argmin(a, axis=None, out=None): >>> np.argmin(b) # Only the first occurrence is returned. 0 + >>> x = np.array([[4,2,3], [1,0,3]]) + >>> index_array = np.argmin(x, axis=-1) + >>> # Same as np.min(x, axis=-1, keepdims=True) + >>> np.take_along_axis(x, np.expand_dims(index_array, axis=-1), axis=-1) + array([[2], + [0]]) + >>> # Same as np.max(x, axis=-1) + >>> np.take_along_axis(x, np.expand_dims(index_array, axis=-1), axis=-1).squeeze(axis=-1) + array([2, 0]) + """ return _wrapfunc(a, 'argmin', axis=axis, out=out) @@ -1404,7 +1449,7 @@ def squeeze(a, axis=None): Raises ------ ValueError - If `axis` is not `None`, and an axis being squeezed is not of length 1 + If `axis` is not None, and an axis being squeezed is not of length 1 See Also -------- @@ -1770,11 +1815,13 @@ def nonzero(a): which returns a row for each non-zero element. .. note:: - When called on a zero-d array or scalar, ``nonzero(a)`` is treated - as ``nonzero(atleast1d(a))``. - ..deprecated:: 1.17.0 - Use `atleast1d` explicitly if this behavior is deliberate. + When called on a zero-d array or scalar, ``nonzero(a)`` is treated + as ``nonzero(atleast1d(a))``. + + .. deprecated:: 1.17.0 + + Use `atleast1d` explicitly if this behavior is deliberate. Parameters ---------- @@ -1938,7 +1985,7 @@ def compress(condition, a, axis=None, out=None): take, choose, diag, diagonal, select ndarray.compress : Equivalent method in ndarray np.extract: Equivalent method when working on 1-D arrays - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Examples -------- @@ -1988,14 +2035,14 @@ def clip(a, a_min, a_max, out=None, **kwargs): ---------- a : array_like Array containing elements to clip. - a_min : scalar or array_like or `None` - Minimum value. If `None`, clipping is not performed on lower + a_min : scalar or array_like or None + Minimum value. If None, clipping is not performed on lower interval edge. Not more than one of `a_min` and `a_max` may be - `None`. - a_max : scalar or array_like or `None` - Maximum value. If `None`, clipping is not performed on upper + None. + a_max : scalar or array_like or None + Maximum value. If None, clipping is not performed on upper interval edge. Not more than one of `a_min` and `a_max` may be - `None`. If `a_min` or `a_max` are array_like, then the three + None. If `a_min` or `a_max` are array_like, then the three arrays will be broadcasted to match their shapes. out : ndarray, optional The results will be placed in this array. It may be the input @@ -2016,7 +2063,7 @@ def clip(a, a_min, a_max, out=None, **kwargs): See Also -------- - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Examples -------- @@ -2199,7 +2246,7 @@ def any(a, axis=None, out=None, keepdims=np._NoValue): Input array or object that can be converted to an array. axis : None or int or tuple of ints, optional Axis or axes along which a logical OR reduction is performed. - The default (`axis` = `None`) is to perform a logical OR over all + The default (``axis=None``) is to perform a logical OR over all the dimensions of the input array. `axis` may be negative, in which case it counts from the last to the first axis. @@ -2212,7 +2259,7 @@ def any(a, axis=None, out=None, keepdims=np._NoValue): the same shape as the expected output and its type is preserved (e.g., if it is of type float, then it will remain so, returning 1.0 for True and 0.0 for False, regardless of the type of `a`). - See `doc.ufuncs` (Section "Output arguments") for details. + See `ufuncs-output-type` for more details. keepdims : bool, optional If this is set to True, the axes which are reduced are left @@ -2285,7 +2332,7 @@ def all(a, axis=None, out=None, keepdims=np._NoValue): Input array or object that can be converted to an array. axis : None or int or tuple of ints, optional Axis or axes along which a logical AND reduction is performed. - The default (`axis` = `None`) is to perform a logical AND over all + The default (``axis=None``) is to perform a logical AND over all the dimensions of the input array. `axis` may be negative, in which case it counts from the last to the first axis. @@ -2297,8 +2344,8 @@ def all(a, axis=None, out=None, keepdims=np._NoValue): Alternate output array in which to place the result. It must have the same shape as the expected output and its type is preserved (e.g., if ``dtype(out)`` is float, the result - will consist of 0.0's and 1.0's). See `doc.ufuncs` (Section - "Output arguments") for more details. + will consist of 0.0's and 1.0's). See `ufuncs-output-type` for more + details. keepdims : bool, optional If this is set to True, the axes which are reduced are left @@ -2376,8 +2423,8 @@ def cumsum(a, axis=None, dtype=None, out=None): out : ndarray, optional Alternative output array in which to place the result. It must have the same shape and buffer length as the expected output - but the type will be cast if necessary. See `doc.ufuncs` - (Section "Output arguments") for more details. + but the type will be cast if necessary. See `ufuncs-output-type` for + more details. Returns ------- @@ -2522,7 +2569,7 @@ def amax(a, axis=None, out=None, keepdims=np._NoValue, initial=np._NoValue, out : ndarray, optional Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. - See `doc.ufuncs` (Section "Output arguments") for more details. + See `ufuncs-output-type` for more details. keepdims : bool, optional If this is set to True, the axes which are reduced are left @@ -2647,7 +2694,7 @@ def amin(a, axis=None, out=None, keepdims=np._NoValue, initial=np._NoValue, out : ndarray, optional Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. - See `doc.ufuncs` (Section "Output arguments") for more details. + See `ufuncs-output-type` for more details. keepdims : bool, optional If this is set to True, the axes which are reduced are left @@ -2778,6 +2825,10 @@ def alen(a): 7 """ + # NumPy 1.18.0, 2019-08-02 + warnings.warn( + "`np.alen` is deprecated, use `len` instead", + DeprecationWarning, stacklevel=2) try: return len(a) except TypeError: @@ -2850,7 +2901,7 @@ def prod(a, axis=None, dtype=None, out=None, keepdims=np._NoValue, See Also -------- ndarray.prod : equivalent method - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Notes ----- @@ -2946,7 +2997,7 @@ def cumprod(a, axis=None, dtype=None, out=None): See Also -------- - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Notes ----- @@ -3092,8 +3143,8 @@ def around(a, decimals=0, out=None): out : ndarray, optional Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output - values will be cast if necessary. See `doc.ufuncs` (Section - "Output arguments") for details. + values will be cast if necessary. See `ufuncs-output-type` for more + details. Returns ------- @@ -3116,10 +3167,37 @@ def around(a, decimals=0, out=None): ----- For values exactly halfway between rounded decimal values, NumPy rounds to the nearest even value. Thus 1.5 and 2.5 round to 2.0, - -0.5 and 0.5 round to 0.0, etc. Results may also be surprising due - to the inexact representation of decimal fractions in the IEEE - floating point standard [1]_ and errors introduced when scaling - by powers of ten. + -0.5 and 0.5 round to 0.0, etc. + + ``np.around`` uses a fast but sometimes inexact algorithm to round + floating-point datatypes. For positive `decimals` it is equivalent to + ``np.true_divide(np.rint(a * 10**decimals), 10**decimals)``, which has + error due to the inexact representation of decimal fractions in the IEEE + floating point standard [1]_ and errors introduced when scaling by powers + of ten. For instance, note the extra "1" in the following: + + >>> np.round(56294995342131.5, 3) + 56294995342131.51 + + If your goal is to print such values with a fixed number of decimals, it is + preferable to use numpy's float printing routines to limit the number of + printed decimals: + + >>> np.format_float_positional(56294995342131.5, precision=3) + '56294995342131.5' + + The float printing routines use an accurate but much more computationally + demanding algorithm to compute the number of digits after the decimal + point. + + Alternatively, Python's builtin `round` function uses a more accurate + but slower algorithm for 64-bit floating point values: + + >>> round(56294995342131.5, 3) + 56294995342131.5 + >>> np.round(16.055, 2), round(16.055, 2) # equals 16.0549999999999997 + (16.06, 16.05) + References ---------- @@ -3180,7 +3258,7 @@ def mean(a, axis=None, dtype=None, out=None, keepdims=np._NoValue): Alternate output array in which to place the result. The default is ``None``; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. - See `doc.ufuncs` for details. + See `ufuncs-output-type` for more details. keepdims : bool, optional If this is set to True, the axes which are reduced are left @@ -3315,7 +3393,7 @@ def std(a, axis=None, dtype=None, out=None, ddof=0, keepdims=np._NoValue): See Also -------- var, mean, nanmean, nanstd, nanvar - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Notes ----- @@ -3410,7 +3488,7 @@ def var(a, axis=None, dtype=None, out=None, ddof=0, keepdims=np._NoValue): instead of a single axis or all the axes as before. dtype : data-type, optional Type to use in computing the variance. For arrays of integer type - the default is `float32`; for arrays of float types it is the same as + the default is `float64`; for arrays of float types it is the same as the array type. out : ndarray, optional Alternate output array in which to place the result. It must have @@ -3440,7 +3518,7 @@ def var(a, axis=None, dtype=None, out=None, ddof=0, keepdims=np._NoValue): See Also -------- std, mean, nanmean, nanstd, nanvar - numpy.doc.ufuncs : Section "Output arguments" + ufuncs-output-type Notes ----- @@ -3569,30 +3647,3 @@ def alltrue(*args, **kwargs): numpy.all : Equivalent function; see for details. """ return all(*args, **kwargs) - - -@array_function_dispatch(_ndim_dispatcher) -def rank(a): - """ - Return the number of dimensions of an array. - - .. note:: - This function is deprecated in NumPy 1.9 to avoid confusion with - `numpy.linalg.matrix_rank`. The ``ndim`` attribute or function - should be used instead. - - See Also - -------- - ndim : equivalent non-deprecated function - - Notes - ----- - In the old Numeric package, `rank` was the term used for the number of - dimensions, but in NumPy `ndim` is used instead. - """ - # 2014-04-12, 1.9 - warnings.warn( - "`rank` is deprecated; use the `ndim` attribute or function instead. " - "To find the rank of a matrix see `numpy.linalg.matrix_rank`.", - VisibleDeprecationWarning, stacklevel=3) - return ndim(a) diff --git a/numpy/core/function_base.py b/numpy/core/function_base.py index c1067299d..538ac8b84 100644 --- a/numpy/core/function_base.py +++ b/numpy/core/function_base.py @@ -3,6 +3,7 @@ from __future__ import division, absolute_import, print_function import functools import warnings import operator +import types from . import numeric as _nx from .numeric import (result_type, NaN, shares_memory, MAY_SHARE_BOUNDS, @@ -17,18 +18,6 @@ array_function_dispatch = functools.partial( overrides.array_function_dispatch, module='numpy') -def _index_deprecate(i, stacklevel=2): - try: - i = operator.index(i) - except TypeError: - msg = ("object of type {} cannot be safely interpreted as " - "an integer.".format(type(i))) - i = int(i) - stacklevel += 1 - warnings.warn(msg, DeprecationWarning, stacklevel=stacklevel) - return i - - def _linspace_dispatcher(start, stop, num=None, endpoint=None, retstep=None, dtype=None, axis=None): return (start, stop) @@ -124,8 +113,13 @@ def linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None, >>> plt.show() """ - # 2016-02-25, 1.12 - num = _index_deprecate(num) + try: + num = operator.index(num) + except TypeError: + raise TypeError( + "object of type {} cannot be safely interpreted as an integer." + .format(type(num))) + if num < 0: raise ValueError("Number of samples, %s, must be non-negative." % num) div = (num - 1) if endpoint else num @@ -145,7 +139,7 @@ def linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None, # from overriding what class is produced, and thus prevents, e.g. use of Quantities, # see gh-7142. Hence, we multiply in place only for standard scalar types. _mult_inplace = _nx.isscalar(delta) - if num > 1: + if div > 0: step = delta / div if _nx.any(step == 0): # Special handling for denormal numbers, gh-5437 @@ -160,7 +154,8 @@ def linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None, else: y = y * step else: - # 0 and 1 item long sequences have an undefined step + # sequences with 0 items or 1 item with endpoint=True (i.e. div <= 0) + # have an undefined step step = NaN # Multiply with delta to allow possible override of output class. y = y * delta @@ -430,28 +425,66 @@ def geomspace(start, stop, num=50, endpoint=True, dtype=None, axis=0): return result.astype(dtype, copy=False) -#always succeed -def _add_docstring(obj, doc): +def _needs_add_docstring(obj): + """ + Returns true if the only way to set the docstring of `obj` from python is + via add_docstring. + + This function errs on the side of being overly conservative. + """ + Py_TPFLAGS_HEAPTYPE = 1 << 9 + + if isinstance(obj, (types.FunctionType, types.MethodType, property)): + return False + + if isinstance(obj, type) and obj.__flags__ & Py_TPFLAGS_HEAPTYPE: + return False + + return True + + +def _add_docstring(obj, doc, warn_on_python): + if warn_on_python and not _needs_add_docstring(obj): + warnings.warn( + "add_newdoc was used on a pure-python object {}. " + "Prefer to attach it directly to the source." + .format(obj), + UserWarning, + stacklevel=3) try: add_docstring(obj, doc) except Exception: pass -def add_newdoc(place, obj, doc): +def add_newdoc(place, obj, doc, warn_on_python=True): """ - Adds documentation to obj which is in module place. + Add documentation to an existing object, typically one defined in C - If doc is a string add it to obj as a docstring + The purpose is to allow easier editing of the docstrings without requiring + a re-compile. This exists primarily for internal use within numpy itself. - If doc is a tuple, then the first element is interpreted as - an attribute of obj and the second as the docstring - (method, docstring) - - If doc is a list, then each element of the list should be a - sequence of length two --> [(method1, docstring1), - (method2, docstring2), ...] + Parameters + ---------- + place : str + The absolute name of the module to import from + obj : str + The name of the object to add documentation to, typically a class or + function name + doc : {str, Tuple[str, str], List[Tuple[str, str]]} + If a string, the documentation to apply to `obj` + + If a tuple, then the first element is interpreted as an attribute of + `obj` and the second as the docstring to apply - ``(method, docstring)`` + + If a list, then each element of the list should be a tuple of length + two - ``[(method1, docstring1), (method2, docstring2), ...]`` + warn_on_python : bool + If True, the default, emit `UserWarning` if this is used to attach + documentation to a pure-python object. + Notes + ----- This routine never raises an error if the docstring can't be written, but will raise an error if the object being documented does not exist. @@ -459,12 +492,23 @@ def add_newdoc(place, obj, doc): in new-style classes or built-in functions. Because this routine never raises an error the caller must check manually that the docstrings were changed. + + Since this function grabs the ``char *`` from a c-level str object and puts + it into the ``tp_doc`` slot of the type of `obj`, it violates a number of + C-API best-practices, by: + + - modifying a `PyTypeObject` after calling `PyType_Ready` + - calling `Py_INCREF` on the str and losing the reference, so the str + will never be released + + If possible it should be avoided. """ new = getattr(__import__(place, globals(), {}, [obj]), obj) if isinstance(doc, str): - _add_docstring(new, doc.strip()) + _add_docstring(new, doc.strip(), warn_on_python) elif isinstance(doc, tuple): - _add_docstring(getattr(new, doc[0]), doc[1].strip()) + attr, docstring = doc + _add_docstring(getattr(new, attr), docstring.strip(), warn_on_python) elif isinstance(doc, list): - for val in doc: - _add_docstring(getattr(new, val[0]), val[1].strip()) + for attr, docstring in doc: + _add_docstring(getattr(new, attr), docstring.strip(), warn_on_python) diff --git a/numpy/core/include/numpy/ndarraytypes.h b/numpy/core/include/numpy/ndarraytypes.h index 1221aeece..ad98d562b 100644 --- a/numpy/core/include/numpy/ndarraytypes.h +++ b/numpy/core/include/numpy/ndarraytypes.h @@ -1095,7 +1095,8 @@ typedef struct PyArrayIterObject_tag PyArrayIterObject; * type of the function which translates a set of coordinates to a * pointer to the data */ -typedef char* (*npy_iter_get_dataptr_t)(PyArrayIterObject* iter, npy_intp*); +typedef char* (*npy_iter_get_dataptr_t)( + PyArrayIterObject* iter, const npy_intp*); struct PyArrayIterObject_tag { PyObject_HEAD @@ -1695,7 +1696,8 @@ PyArray_CLEARFLAGS(PyArrayObject *arr, int flags) #define PyDataType_ISOBJECT(obj) PyTypeNum_ISOBJECT(((PyArray_Descr*)(obj))->type_num) #define PyDataType_HASFIELDS(obj) (((PyArray_Descr *)(obj))->names != NULL) #define PyDataType_HASSUBARRAY(dtype) ((dtype)->subarray != NULL) -#define PyDataType_ISUNSIZED(dtype) ((dtype)->elsize == 0) +#define PyDataType_ISUNSIZED(dtype) ((dtype)->elsize == 0 && \ + !PyDataType_HASFIELDS(dtype)) #define PyDataType_MAKEUNSIZED(dtype) ((dtype)->elsize = 0) #define PyArray_ISBOOL(obj) PyTypeNum_ISBOOL(PyArray_TYPE(obj)) diff --git a/numpy/core/include/numpy/npy_common.h b/numpy/core/include/numpy/npy_common.h index 108c0a202..27b83f7b5 100644 --- a/numpy/core/include/numpy/npy_common.h +++ b/numpy/core/include/numpy/npy_common.h @@ -44,10 +44,14 @@ #else #define NPY_GCC_TARGET_AVX #endif + +#if defined HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS +#define HAVE_ATTRIBUTE_TARGET_FMA +#define NPY_GCC_TARGET_FMA __attribute__((target("avx2,fma"))) +#endif + #if defined HAVE_ATTRIBUTE_TARGET_AVX2 && defined HAVE_LINK_AVX2 #define NPY_GCC_TARGET_AVX2 __attribute__((target("avx2"))) -#elif defined HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS -#define NPY_GCC_TARGET_AVX2 __attribute__((target("avx2"))) #else #define NPY_GCC_TARGET_AVX2 #endif diff --git a/numpy/core/include/numpy/npy_math.h b/numpy/core/include/numpy/npy_math.h index 6a78ff3c2..69e690f28 100644 --- a/numpy/core/include/numpy/npy_math.h +++ b/numpy/core/include/numpy/npy_math.h @@ -144,7 +144,22 @@ NPY_INLINE static float __npy_nzerof(void) #define NPY_COEFF_Q3_LOGf 9.864942958519418960339e-01f #define NPY_COEFF_Q4_LOGf 1.546476374983906719538e-01f #define NPY_COEFF_Q5_LOGf 5.875095403124574342950e-03f - +/* + * Constants used in vector implementation of sinf/cosf(x) + */ +#define NPY_TWO_O_PIf 0x1.45f306p-1f +#define NPY_CODY_WAITE_PI_O_2_HIGHf -0x1.921fb0p+00f +#define NPY_CODY_WAITE_PI_O_2_MEDf -0x1.5110b4p-22f +#define NPY_CODY_WAITE_PI_O_2_LOWf -0x1.846988p-48f +#define NPY_COEFF_INVF0_COSINEf 0x1.000000p+00f +#define NPY_COEFF_INVF2_COSINEf -0x1.000000p-01f +#define NPY_COEFF_INVF4_COSINEf 0x1.55553cp-05f +#define NPY_COEFF_INVF6_COSINEf -0x1.6c06dcp-10f +#define NPY_COEFF_INVF8_COSINEf 0x1.98e616p-16f +#define NPY_COEFF_INVF3_SINEf -0x1.555556p-03f +#define NPY_COEFF_INVF5_SINEf 0x1.11119ap-07f +#define NPY_COEFF_INVF7_SINEf -0x1.a06bbap-13f +#define NPY_COEFF_INVF9_SINEf 0x1.7d3bbcp-19f /* * Integer functions. */ @@ -162,6 +177,37 @@ NPY_INPLACE npy_long npy_lcml(npy_long a, npy_long b); NPY_INPLACE npy_longlong npy_gcdll(npy_longlong a, npy_longlong b); NPY_INPLACE npy_longlong npy_lcmll(npy_longlong a, npy_longlong b); +NPY_INPLACE npy_ubyte npy_rshiftuhh(npy_ubyte a, npy_ubyte b); +NPY_INPLACE npy_ubyte npy_lshiftuhh(npy_ubyte a, npy_ubyte b); +NPY_INPLACE npy_ushort npy_rshiftuh(npy_ushort a, npy_ushort b); +NPY_INPLACE npy_ushort npy_lshiftuh(npy_ushort a, npy_ushort b); +NPY_INPLACE npy_uint npy_rshiftu(npy_uint a, npy_uint b); +NPY_INPLACE npy_uint npy_lshiftu(npy_uint a, npy_uint b); +NPY_INPLACE npy_ulong npy_rshiftul(npy_ulong a, npy_ulong b); +NPY_INPLACE npy_ulong npy_lshiftul(npy_ulong a, npy_ulong b); +NPY_INPLACE npy_ulonglong npy_rshiftull(npy_ulonglong a, npy_ulonglong b); +NPY_INPLACE npy_ulonglong npy_lshiftull(npy_ulonglong a, npy_ulonglong b); + +NPY_INPLACE npy_byte npy_rshifthh(npy_byte a, npy_byte b); +NPY_INPLACE npy_byte npy_lshifthh(npy_byte a, npy_byte b); +NPY_INPLACE npy_short npy_rshifth(npy_short a, npy_short b); +NPY_INPLACE npy_short npy_lshifth(npy_short a, npy_short b); +NPY_INPLACE npy_int npy_rshift(npy_int a, npy_int b); +NPY_INPLACE npy_int npy_lshift(npy_int a, npy_int b); +NPY_INPLACE npy_long npy_rshiftl(npy_long a, npy_long b); +NPY_INPLACE npy_long npy_lshiftl(npy_long a, npy_long b); +NPY_INPLACE npy_longlong npy_rshiftll(npy_longlong a, npy_longlong b); +NPY_INPLACE npy_longlong npy_lshiftll(npy_longlong a, npy_longlong b); + +/* + * avx function has a common API for both sin & cos. This enum is used to + * distinguish between the two + */ +typedef enum { + npy_compute_sin, + npy_compute_cos +} NPY_TRIG_OP; + /* * C99 double math funcs */ diff --git a/numpy/core/include/numpy/numpyconfig.h b/numpy/core/include/numpy/numpyconfig.h index ab198f36b..4bca82f9f 100644 --- a/numpy/core/include/numpy/numpyconfig.h +++ b/numpy/core/include/numpy/numpyconfig.h @@ -37,5 +37,8 @@ #define NPY_1_13_API_VERSION 0x00000008 #define NPY_1_14_API_VERSION 0x00000008 #define NPY_1_15_API_VERSION 0x00000008 +#define NPY_1_16_API_VERSION 0x00000008 +#define NPY_1_17_API_VERSION 0x00000008 +#define NPY_1_18_API_VERSION 0x00000008 #endif diff --git a/numpy/core/include/numpy/random/bitgen.h b/numpy/core/include/numpy/random/bitgen.h new file mode 100644 index 000000000..83c2858dd --- /dev/null +++ b/numpy/core/include/numpy/random/bitgen.h @@ -0,0 +1,20 @@ +#ifndef _RANDOM_BITGEN_H +#define _RANDOM_BITGEN_H + +#pragma once +#include <stddef.h> +#include <stdbool.h> +#include <stdint.h> + +/* Must match the declaration in numpy/random/<any>.pxd */ + +typedef struct bitgen { + void *state; + uint64_t (*next_uint64)(void *st); + uint32_t (*next_uint32)(void *st); + double (*next_double)(void *st); + uint64_t (*next_raw)(void *st); +} bitgen_t; + + +#endif diff --git a/numpy/core/include/numpy/random/distributions.h b/numpy/core/include/numpy/random/distributions.h new file mode 100644 index 000000000..c474c4d14 --- /dev/null +++ b/numpy/core/include/numpy/random/distributions.h @@ -0,0 +1,200 @@ +#ifndef _RANDOMDGEN__DISTRIBUTIONS_H_ +#define _RANDOMDGEN__DISTRIBUTIONS_H_ + +#include "Python.h" +#include "numpy/npy_common.h" +#include <stddef.h> +#include <stdbool.h> +#include <stdint.h> + +#include "numpy/npy_math.h" +#include "numpy/random/bitgen.h" + +/* + * RAND_INT_TYPE is used to share integer generators with RandomState which + * used long in place of int64_t. If changing a distribution that uses + * RAND_INT_TYPE, then the original unmodified copy must be retained for + * use in RandomState by copying to the legacy distributions source file. + */ +#ifdef NP_RANDOM_LEGACY +#define RAND_INT_TYPE long +#define RAND_INT_MAX LONG_MAX +#else +#define RAND_INT_TYPE int64_t +#define RAND_INT_MAX INT64_MAX +#endif + +#ifdef _MSC_VER +#define DECLDIR __declspec(dllexport) +#else +#define DECLDIR extern +#endif + +#ifndef MIN +#define MIN(x, y) (((x) < (y)) ? x : y) +#define MAX(x, y) (((x) > (y)) ? x : y) +#endif + +#ifndef M_PI +#define M_PI 3.14159265358979323846264338328 +#endif + +typedef struct s_binomial_t { + int has_binomial; /* !=0: following parameters initialized for binomial */ + double psave; + RAND_INT_TYPE nsave; + double r; + double q; + double fm; + RAND_INT_TYPE m; + double p1; + double xm; + double xl; + double xr; + double c; + double laml; + double lamr; + double p2; + double p3; + double p4; +} binomial_t; + +DECLDIR float random_standard_uniform_f(bitgen_t *bitgen_state); +DECLDIR double random_standard_uniform(bitgen_t *bitgen_state); +DECLDIR void random_standard_uniform_fill(bitgen_t *, npy_intp, double *); +DECLDIR void random_standard_uniform_fill_f(bitgen_t *, npy_intp, float *); + +DECLDIR int64_t random_positive_int64(bitgen_t *bitgen_state); +DECLDIR int32_t random_positive_int32(bitgen_t *bitgen_state); +DECLDIR int64_t random_positive_int(bitgen_t *bitgen_state); +DECLDIR uint64_t random_uint(bitgen_t *bitgen_state); + +DECLDIR double random_standard_exponential(bitgen_t *bitgen_state); +DECLDIR float random_standard_exponential_f(bitgen_t *bitgen_state); +DECLDIR void random_standard_exponential_fill(bitgen_t *, npy_intp, double *); +DECLDIR void random_standard_exponential_fill_f(bitgen_t *, npy_intp, float *); +DECLDIR void random_standard_exponential_inv_fill(bitgen_t *, npy_intp, double *); +DECLDIR void random_standard_exponential_inv_fill_f(bitgen_t *, npy_intp, float *); + +DECLDIR double random_standard_normal(bitgen_t *bitgen_state); +DECLDIR float random_standard_normal_f(bitgen_t *bitgen_state); +DECLDIR void random_standard_normal_fill(bitgen_t *, npy_intp, double *); +DECLDIR void random_standard_normal_fill_f(bitgen_t *, npy_intp, float *); +DECLDIR double random_standard_gamma(bitgen_t *bitgen_state, double shape); +DECLDIR float random_standard_gamma_f(bitgen_t *bitgen_state, float shape); + +DECLDIR double random_normal(bitgen_t *bitgen_state, double loc, double scale); + +DECLDIR double random_gamma(bitgen_t *bitgen_state, double shape, double scale); +DECLDIR float random_gamma_f(bitgen_t *bitgen_state, float shape, float scale); + +DECLDIR double random_exponential(bitgen_t *bitgen_state, double scale); +DECLDIR double random_uniform(bitgen_t *bitgen_state, double lower, double range); +DECLDIR double random_beta(bitgen_t *bitgen_state, double a, double b); +DECLDIR double random_chisquare(bitgen_t *bitgen_state, double df); +DECLDIR double random_f(bitgen_t *bitgen_state, double dfnum, double dfden); +DECLDIR double random_standard_cauchy(bitgen_t *bitgen_state); +DECLDIR double random_pareto(bitgen_t *bitgen_state, double a); +DECLDIR double random_weibull(bitgen_t *bitgen_state, double a); +DECLDIR double random_power(bitgen_t *bitgen_state, double a); +DECLDIR double random_laplace(bitgen_t *bitgen_state, double loc, double scale); +DECLDIR double random_gumbel(bitgen_t *bitgen_state, double loc, double scale); +DECLDIR double random_logistic(bitgen_t *bitgen_state, double loc, double scale); +DECLDIR double random_lognormal(bitgen_t *bitgen_state, double mean, double sigma); +DECLDIR double random_rayleigh(bitgen_t *bitgen_state, double mode); +DECLDIR double random_standard_t(bitgen_t *bitgen_state, double df); +DECLDIR double random_noncentral_chisquare(bitgen_t *bitgen_state, double df, + double nonc); +DECLDIR double random_noncentral_f(bitgen_t *bitgen_state, double dfnum, + double dfden, double nonc); +DECLDIR double random_wald(bitgen_t *bitgen_state, double mean, double scale); +DECLDIR double random_vonmises(bitgen_t *bitgen_state, double mu, double kappa); +DECLDIR double random_triangular(bitgen_t *bitgen_state, double left, double mode, + double right); + +DECLDIR RAND_INT_TYPE random_poisson(bitgen_t *bitgen_state, double lam); +DECLDIR RAND_INT_TYPE random_negative_binomial(bitgen_t *bitgen_state, double n, + double p); + +DECLDIR int64_t random_binomial(bitgen_t *bitgen_state, double p, + int64_t n, binomial_t *binomial); + +DECLDIR RAND_INT_TYPE random_logseries(bitgen_t *bitgen_state, double p); +DECLDIR RAND_INT_TYPE random_geometric(bitgen_t *bitgen_state, double p); +DECLDIR RAND_INT_TYPE random_zipf(bitgen_t *bitgen_state, double a); +DECLDIR int64_t random_hypergeometric(bitgen_t *bitgen_state, + int64_t good, int64_t bad, int64_t sample); +DECLDIR uint64_t random_interval(bitgen_t *bitgen_state, uint64_t max); + +/* Generate random uint64 numbers in closed interval [off, off + rng]. */ +DECLDIR uint64_t random_bounded_uint64(bitgen_t *bitgen_state, uint64_t off, + uint64_t rng, uint64_t mask, + bool use_masked); + +/* Generate random uint32 numbers in closed interval [off, off + rng]. */ +DECLDIR uint32_t random_buffered_bounded_uint32(bitgen_t *bitgen_state, + uint32_t off, uint32_t rng, + uint32_t mask, bool use_masked, + int *bcnt, uint32_t *buf); +DECLDIR uint16_t random_buffered_bounded_uint16(bitgen_t *bitgen_state, + uint16_t off, uint16_t rng, + uint16_t mask, bool use_masked, + int *bcnt, uint32_t *buf); +DECLDIR uint8_t random_buffered_bounded_uint8(bitgen_t *bitgen_state, uint8_t off, + uint8_t rng, uint8_t mask, + bool use_masked, int *bcnt, + uint32_t *buf); +DECLDIR npy_bool random_buffered_bounded_bool(bitgen_t *bitgen_state, npy_bool off, + npy_bool rng, npy_bool mask, + bool use_masked, int *bcnt, + uint32_t *buf); + +DECLDIR void random_bounded_uint64_fill(bitgen_t *bitgen_state, uint64_t off, + uint64_t rng, npy_intp cnt, + bool use_masked, uint64_t *out); +DECLDIR void random_bounded_uint32_fill(bitgen_t *bitgen_state, uint32_t off, + uint32_t rng, npy_intp cnt, + bool use_masked, uint32_t *out); +DECLDIR void random_bounded_uint16_fill(bitgen_t *bitgen_state, uint16_t off, + uint16_t rng, npy_intp cnt, + bool use_masked, uint16_t *out); +DECLDIR void random_bounded_uint8_fill(bitgen_t *bitgen_state, uint8_t off, + uint8_t rng, npy_intp cnt, + bool use_masked, uint8_t *out); +DECLDIR void random_bounded_bool_fill(bitgen_t *bitgen_state, npy_bool off, + npy_bool rng, npy_intp cnt, + bool use_masked, npy_bool *out); + +DECLDIR void random_multinomial(bitgen_t *bitgen_state, RAND_INT_TYPE n, RAND_INT_TYPE *mnix, + double *pix, npy_intp d, binomial_t *binomial); + +/* multivariate hypergeometric, "count" method */ +DECLDIR int random_multivariate_hypergeometric_count(bitgen_t *bitgen_state, + int64_t total, + size_t num_colors, int64_t *colors, + int64_t nsample, + size_t num_variates, int64_t *variates); + +/* multivariate hypergeometric, "marginals" method */ +DECLDIR void random_multivariate_hypergeometric_marginals(bitgen_t *bitgen_state, + int64_t total, + size_t num_colors, int64_t *colors, + int64_t nsample, + size_t num_variates, int64_t *variates); + +/* Common to legacy-distributions.c and distributions.c but not exported */ + +RAND_INT_TYPE random_binomial_btpe(bitgen_t *bitgen_state, + RAND_INT_TYPE n, + double p, + binomial_t *binomial); +RAND_INT_TYPE random_binomial_inversion(bitgen_t *bitgen_state, + RAND_INT_TYPE n, + double p, + binomial_t *binomial); +double random_loggam(double x); +static NPY_INLINE double next_double(bitgen_t *bitgen_state) { + return bitgen_state->next_double(bitgen_state->state); +} + +#endif diff --git a/numpy/core/include/numpy/ufuncobject.h b/numpy/core/include/numpy/ufuncobject.h index 15dcdf010..5ff4a0041 100644 --- a/numpy/core/include/numpy/ufuncobject.h +++ b/numpy/core/include/numpy/ufuncobject.h @@ -340,14 +340,6 @@ typedef struct _loop1d_info { #define UFUNC_PYVALS_NAME "UFUNC_PYVALS" -#define UFUNC_CHECK_ERROR(arg) \ - do {if ((((arg)->obj & UFUNC_OBJ_NEEDS_API) && PyErr_Occurred()) || \ - ((arg)->errormask && \ - PyUFunc_checkfperr((arg)->errormask, \ - (arg)->errobj, \ - &(arg)->first))) \ - goto fail;} while (0) - /* * THESE MACROS ARE DEPRECATED. * Use npy_set_floatstatus_* in the npymath library. @@ -357,10 +349,6 @@ typedef struct _loop1d_info { #define UFUNC_FPE_UNDERFLOW NPY_FPE_UNDERFLOW #define UFUNC_FPE_INVALID NPY_FPE_INVALID -#define UFUNC_CHECK_STATUS(ret) \ - { \ - ret = npy_clear_floatstatus(); \ - } #define generate_divbyzero_error() npy_set_floatstatus_divbyzero() #define generate_overflow_error() npy_set_floatstatus_overflow() diff --git a/numpy/core/info.py b/numpy/core/info.py deleted file mode 100644 index c6f7bbcf2..000000000 --- a/numpy/core/info.py +++ /dev/null @@ -1,87 +0,0 @@ -"""Defines a multi-dimensional array and useful procedures for Numerical computation. - -Functions - -- array - NumPy Array construction -- zeros - Return an array of all zeros -- empty - Return an uninitialized array -- shape - Return shape of sequence or array -- rank - Return number of dimensions -- size - Return number of elements in entire array or a - certain dimension -- fromstring - Construct array from (byte) string -- take - Select sub-arrays using sequence of indices -- put - Set sub-arrays using sequence of 1-D indices -- putmask - Set portion of arrays using a mask -- reshape - Return array with new shape -- repeat - Repeat elements of array -- choose - Construct new array from indexed array tuple -- correlate - Correlate two 1-d arrays -- searchsorted - Search for element in 1-d array -- sum - Total sum over a specified dimension -- average - Average, possibly weighted, over axis or array. -- cumsum - Cumulative sum over a specified dimension -- product - Total product over a specified dimension -- cumproduct - Cumulative product over a specified dimension -- alltrue - Logical and over an entire axis -- sometrue - Logical or over an entire axis -- allclose - Tests if sequences are essentially equal - -More Functions: - -- arange - Return regularly spaced array -- asarray - Guarantee NumPy array -- convolve - Convolve two 1-d arrays -- swapaxes - Exchange axes -- concatenate - Join arrays together -- transpose - Permute axes -- sort - Sort elements of array -- argsort - Indices of sorted array -- argmax - Index of largest value -- argmin - Index of smallest value -- inner - Innerproduct of two arrays -- dot - Dot product (matrix multiplication) -- outer - Outerproduct of two arrays -- resize - Return array with arbitrary new shape -- indices - Tuple of indices -- fromfunction - Construct array from universal function -- diagonal - Return diagonal array -- trace - Trace of array -- dump - Dump array to file object (pickle) -- dumps - Return pickled string representing data -- load - Return array stored in file object -- loads - Return array from pickled string -- ravel - Return array as 1-D -- nonzero - Indices of nonzero elements for 1-D array -- shape - Shape of array -- where - Construct array from binary result -- compress - Elements of array where condition is true -- clip - Clip array between two values -- ones - Array of all ones -- identity - 2-D identity array (matrix) - -(Universal) Math Functions - - add logical_or exp - subtract logical_xor log - multiply logical_not log10 - divide maximum sin - divide_safe minimum sinh - conjugate bitwise_and sqrt - power bitwise_or tan - absolute bitwise_xor tanh - negative invert ceil - greater left_shift fabs - greater_equal right_shift floor - less arccos arctan2 - less_equal arcsin fmod - equal arctan hypot - not_equal cos around - logical_and cosh sign - arccosh arcsinh arctanh - -""" -from __future__ import division, absolute_import, print_function - -depends = ['testing'] -global_symbols = ['*'] diff --git a/numpy/core/numeric.py b/numpy/core/numeric.py index ea2ef900e..1e011e2e7 100644 --- a/numpy/core/numeric.py +++ b/numpy/core/numeric.py @@ -26,6 +26,7 @@ if sys.version_info[0] < 3: from . import overrides from . import umath +from . import shape_base from .overrides import set_module from .umath import (multiply, invert, sin, PINF, NAN) from . import numerictypes @@ -48,14 +49,6 @@ array_function_dispatch = functools.partial( overrides.array_function_dispatch, module='numpy') -def loads(*args, **kwargs): - # NumPy 1.15.0, 2017-12-10 - warnings.warn( - "np.core.numeric.loads is deprecated, use pickle.loads instead", - DeprecationWarning, stacklevel=2) - return pickle.loads(*args, **kwargs) - - __all__ = [ 'newaxis', 'ndarray', 'flatiter', 'nditer', 'nested_iters', 'ufunc', 'arange', 'array', 'zeros', 'count_nonzero', 'empty', 'broadcast', 'dtype', @@ -66,7 +59,7 @@ __all__ = [ 'correlate', 'convolve', 'inner', 'dot', 'outer', 'vdot', 'roll', 'rollaxis', 'moveaxis', 'cross', 'tensordot', 'little_endian', 'fromiter', 'array_equal', 'array_equiv', 'indices', 'fromfunction', - 'isclose', 'load', 'loads', 'isscalar', 'binary_repr', 'base_repr', 'ones', + 'isclose', 'isscalar', 'binary_repr', 'base_repr', 'ones', 'identity', 'allclose', 'compare_chararrays', 'putmask', 'flatnonzero', 'Inf', 'inf', 'infty', 'Infinity', 'nan', 'NaN', 'False_', 'True_', 'bitwise_not', 'CLIP', 'RAISE', 'WRAP', 'MAXDIMS', @@ -299,7 +292,7 @@ def full(shape, fill_value, dtype=None, order='C'): fill_value : scalar Fill value. dtype : data-type, optional - The desired data-type for the array The default, `None`, means + The desired data-type for the array The default, None, means `np.array(fill_value).dtype`. order : {'C', 'F'}, optional Whether to store multidimensional data in C- or Fortran-contiguous @@ -530,7 +523,7 @@ def isfortran(a): C-ordered arrays evaluate as False even if they are also FORTRAN-ordered. - >>> np.isfortran(np.array([1, 2], order='FORTRAN')) + >>> np.isfortran(np.array([1, 2], order='F')) False """ @@ -553,8 +546,10 @@ def argwhere(a): Returns ------- - index_array : ndarray + index_array : (N, a.ndim) ndarray Indices of elements that are non-zero. Indices are grouped by element. + This array will have shape ``(N, a.ndim)`` where ``N`` is the number of + non-zero items. See Also -------- @@ -562,7 +557,8 @@ def argwhere(a): Notes ----- - ``np.argwhere(a)`` is the same as ``np.transpose(np.nonzero(a))``. + ``np.argwhere(a)`` is almost the same as ``np.transpose(np.nonzero(a))``, + but produces a result of the correct shape for a 0D array. The output of ``argwhere`` is not suitable for indexing arrays. For this purpose use ``nonzero(a)`` instead. @@ -580,6 +576,11 @@ def argwhere(a): [1, 2]]) """ + # nonzero does not behave well on 0d, so promote to 1d + if np.ndim(a) == 0: + a = shape_base.atleast_1d(a) + # then remove the added dimension + return argwhere(a)[:,:0] return transpose(nonzero(a)) @@ -937,7 +938,7 @@ def tensordot(a, b, axes=2): Returns ------- output : ndarray - The tensor dot product of the input. + The tensor dot product of the input. See Also -------- @@ -959,6 +960,9 @@ def tensordot(a, b, axes=2): two sequences of the same length, with the first axis to sum over given first in both sequences, the second axis second, and so forth. + The shape of the result consists of the non-contracted axes of the + first tensor, followed by the non-contracted axes of the second. + Examples -------- A "traditional" example: @@ -1780,19 +1784,19 @@ def _frombuffer(buf, dtype, shape, order): @set_module('numpy') -def isscalar(num): +def isscalar(element): """ - Returns True if the type of `num` is a scalar type. + Returns True if the type of `element` is a scalar type. Parameters ---------- - num : any + element : any Input argument, can be of any type and shape. Returns ------- val : bool - True if `num` is a scalar type, False if it is not. + True if `element` is a scalar type, False if it is not. See Also -------- @@ -1800,10 +1804,14 @@ def isscalar(num): Notes ----- - In almost all cases ``np.ndim(x) == 0`` should be used instead of this - function, as that will also return true for 0d arrays. This is how - numpy overloads functions in the style of the ``dx`` arguments to `gradient` - and the ``bins`` argument to `histogram`. Some key differences: + If you need a stricter way to identify a *numerical* scalar, use + ``isinstance(x, numbers.Number)``, as that returns ``False`` for most + non-numerical elements such as strings. + + In most cases ``np.ndim(x) == 0`` should be used instead of this function, + as that will also return true for 0d arrays. This is how numpy overloads + functions in the style of the ``dx`` arguments to `gradient` and the ``bins`` + argument to `histogram`. Some key differences: +--------------------------------------+---------------+-------------------+ | x |``isscalar(x)``|``np.ndim(x) == 0``| @@ -1851,9 +1859,9 @@ def isscalar(num): True """ - return (isinstance(num, generic) - or type(num) in ScalarType - or isinstance(num, numbers.Number)) + return (isinstance(element, generic) + or type(element) in ScalarType + or isinstance(element, numbers.Number)) @set_module('numpy') @@ -1935,6 +1943,10 @@ def binary_repr(num, width=None): "will raise an error in the future.", DeprecationWarning, stacklevel=3) + # Ensure that num is a Python integer to avoid overflow or unwanted + # casts to floating point. + num = operator.index(num) + if num == 0: return '0' * (width or 1) @@ -2024,30 +2036,6 @@ def base_repr(number, base=2, padding=0): return ''.join(reversed(res or '0')) -def load(file): - """ - Wrapper around cPickle.load which accepts either a file-like object or - a filename. - - Note that the NumPy binary format is not based on pickle/cPickle anymore. - For details on the preferred way of loading and saving files, see `load` - and `save`. - - See Also - -------- - load, save - - """ - # NumPy 1.15.0, 2017-12-10 - warnings.warn( - "np.core.numeric.load is deprecated, use pickle.load instead", - DeprecationWarning, stacklevel=2) - if isinstance(file, type("")): - with open(file, "rb") as file_pointer: - return pickle.load(file_pointer) - return pickle.load(file) - - # These are all essentially abbreviations # These might wind up in a special abbreviations module @@ -2110,9 +2098,9 @@ def allclose(a, b, rtol=1.e-5, atol=1.e-8, equal_nan=False): `atol` are added together to compare against the absolute difference between `a` and `b`. - If either array contains one or more NaNs, False is returned. - Infs are treated as equal if they are in the same place and of the same - sign in both arrays. + NaNs are treated as equal if they are in the same place and if + ``equal_nan=True``. Infs are treated as equal if they are in the same + place and of the same sign in both arrays. Parameters ---------- diff --git a/numpy/core/numerictypes.py b/numpy/core/numerictypes.py index ab1ff65a4..761c7087c 100644 --- a/numpy/core/numerictypes.py +++ b/numpy/core/numerictypes.py @@ -485,7 +485,7 @@ def sctype2char(sctype): Examples -------- - >>> for sctype in [np.int32, np.double, np.complex, np.string_, np.ndarray]: + >>> for sctype in [np.int32, np.double, np.complex_, np.string_, np.ndarray]: ... print(np.sctype2char(sctype)) l # may vary d diff --git a/numpy/core/overrides.py b/numpy/core/overrides.py index 04a5a995f..55c7bd1ea 100644 --- a/numpy/core/overrides.py +++ b/numpy/core/overrides.py @@ -109,6 +109,18 @@ def set_module(module): return decorator + +# Call textwrap.dedent here instead of in the function so as to avoid +# calling dedent multiple times on the same text +_wrapped_func_source = textwrap.dedent(""" + @functools.wraps(implementation) + def {name}(*args, **kwargs): + relevant_args = dispatcher(*args, **kwargs) + return implement_array_function( + implementation, {name}, relevant_args, args, kwargs) + """) + + def array_function_dispatch(dispatcher, module=None, verify=True, docs_from_dispatcher=False): """Decorator for adding dispatch with the __array_function__ protocol. @@ -163,13 +175,7 @@ def array_function_dispatch(dispatcher, module=None, verify=True, # more interpettable name. Otherwise, the original function does not # show up at all in many cases, e.g., if it's written in C or if the # dispatcher gets an invalid keyword argument. - source = textwrap.dedent(""" - @functools.wraps(implementation) - def {name}(*args, **kwargs): - relevant_args = dispatcher(*args, **kwargs) - return implement_array_function( - implementation, {name}, relevant_args, args, kwargs) - """).format(name=implementation.__name__) + source = _wrapped_func_source.format(name=implementation.__name__) source_object = compile( source, filename='<__array_function__ internals>', mode='exec') diff --git a/numpy/core/records.py b/numpy/core/records.py index 8a5fee541..a1cad9075 100644 --- a/numpy/core/records.py +++ b/numpy/core/records.py @@ -163,7 +163,7 @@ class format_parser(object): self._createdescr(byteorder) self.dtype = self._descr - def _parseFormats(self, formats, aligned=0): + def _parseFormats(self, formats, aligned=False): """ Parse the field formats """ if formats is None: diff --git a/numpy/core/setup.py b/numpy/core/setup.py index 0a2a8233e..974ec4628 100644 --- a/numpy/core/setup.py +++ b/numpy/core/setup.py @@ -394,7 +394,7 @@ def visibility_define(config): def configuration(parent_package='',top_path=None): from numpy.distutils.misc_util import Configuration, dot_join - from numpy.distutils.system_info import get_info + from numpy.distutils.system_info import get_info, dict_append config = Configuration('core', parent_package, top_path) local_dir = config.local_path @@ -463,8 +463,14 @@ def configuration(parent_package='',top_path=None): rep = check_long_double_representation(config_cmd) moredefs.append(('HAVE_LDOUBLE_%s' % rep, 1)) + if check_for_right_shift_internal_compiler_error(config_cmd): + moredefs.append('NPY_DO_NOT_OPTIMIZE_LONG_right_shift') + moredefs.append('NPY_DO_NOT_OPTIMIZE_ULONG_right_shift') + moredefs.append('NPY_DO_NOT_OPTIMIZE_LONGLONG_right_shift') + moredefs.append('NPY_DO_NOT_OPTIMIZE_ULONGLONG_right_shift') + # Py3K check - if sys.version_info[0] == 3: + if sys.version_info[0] >= 3: moredefs.append(('NPY_PY3K', 1)) # Generate the config.h file from moredefs @@ -491,10 +497,10 @@ def configuration(parent_package='',top_path=None): #endif """)) - print('File:', target) + log.info('File: %s' % target) with open(target) as target_f: - print(target_f.read()) - print('EOF') + log.info(target_f.read()) + log.info('EOF') else: mathlibs = [] with open(target) as target_f: @@ -581,10 +587,10 @@ def configuration(parent_package='',top_path=None): """)) # Dump the numpyconfig.h header to stdout - print('File: %s' % target) + log.info('File: %s' % target) with open(target) as target_f: - print(target_f.read()) - print('EOF') + log.info(target_f.read()) + log.info('EOF') config.add_data_files((header_dir, target)) return target @@ -633,23 +639,6 @@ def configuration(parent_package='',top_path=None): ] ####################################################################### - # dummy module # - ####################################################################### - - # npymath needs the config.h and numpyconfig.h files to be generated, but - # build_clib cannot handle generate_config_h and generate_numpyconfig_h - # (don't ask). Because clib are generated before extensions, we have to - # explicitly add an extension which has generate_config_h and - # generate_numpyconfig_h as sources *before* adding npymath. - - config.add_extension('_dummy', - sources=[join('src', 'dummymodule.c'), - generate_config_h, - generate_numpyconfig_h, - generate_numpy_api] - ) - - ####################################################################### # npymath library # ####################################################################### @@ -666,6 +655,9 @@ def configuration(parent_package='',top_path=None): # compiler does not work). st = config_cmd.try_link('int main(void) { return 0;}') if not st: + # rerun the failing command in verbose mode + config_cmd.compiler.verbose = True + config_cmd.try_link('int main(void) { return 0;}') raise RuntimeError("Broken toolchain: cannot link a simple C program") mlibs = check_mathlib(config_cmd) @@ -761,8 +753,14 @@ def configuration(parent_package='',top_path=None): join('src', 'common', 'numpyos.c'), ] - blas_info = get_info('blas_opt', 0) - if blas_info and ('HAVE_CBLAS', None) in blas_info.get('define_macros', []): + if os.environ.get('NPY_USE_BLAS_ILP64', "0") != "0": + blas_info = get_info('blas_ilp64_opt', 2) + else: + blas_info = get_info('blas_opt', 0) + + have_blas = blas_info and ('HAVE_CBLAS', None) in blas_info.get('define_macros', []) + + if have_blas: extra_info = blas_info # These files are also in MANIFEST.in so that they are always in # the source distribution independently of HAVE_CBLAS. diff --git a/numpy/core/setup_common.py b/numpy/core/setup_common.py index 80706c99f..6356f08ba 100644 --- a/numpy/core/setup_common.py +++ b/numpy/core/setup_common.py @@ -5,6 +5,7 @@ import sys import warnings import copy import binascii +import textwrap from numpy.distutils.misc_util import mingw32 @@ -14,7 +15,7 @@ from numpy.distutils.misc_util import mingw32 #------------------- # How to change C_API_VERSION ? # - increase C_API_VERSION value -# - record the hash for the new C API with the script cversions.py +# - record the hash for the new C API with the cversions.py script # and add the hash to cversions.txt # The hash values are used to remind developers when the C API number was not # updated - generates a MismatchCAPIWarning warning which is turned into an @@ -88,14 +89,13 @@ def check_api_version(apiversion, codegen_dir): # codegen_dir have been updated without the API version being # updated. Any modification in those .txt files should be reflected # in the api and eventually abi versions. - # To compute the checksum of the current API, use - # code_generators/cversions.py script + # To compute the checksum of the current API, use numpy/core/cversions.py if not curapi_hash == api_hash: msg = ("API mismatch detected, the C API version " "numbers have to be updated. Current C api version is %d, " - "with checksum %s, but recorded checksum for C API version %d in " - "codegen_dir/cversions.txt is %s. If functions were added in the " - "C API, you have to update C_API_VERSION in %s." + "with checksum %s, but recorded checksum for C API version %d " + "in core/codegen_dir/cversions.txt is %s. If functions were " + "added in the C API, you have to update C_API_VERSION in %s." ) warnings.warn(msg % (apiversion, curapi_hash, apiversion, api_hash, __file__), @@ -179,9 +179,10 @@ OPTIONAL_FUNCTION_ATTRIBUTES = [('__attribute__((optimize("unroll-loops")))', # gcc 4.8.4 support attributes but not with intrisics # tested via "#include<%s> int %s %s(void *){code; return 0;};" % (header, attribute, name, code) # function name will be converted to HAVE_<upper-case-name> preprocessor macro -OPTIONAL_FUNCTION_ATTRIBUTES_WITH_INTRINSICS = [('__attribute__((target("avx2")))', +OPTIONAL_FUNCTION_ATTRIBUTES_WITH_INTRINSICS = [('__attribute__((target("avx2,fma")))', 'attribute_target_avx2_with_intrinsics', - '__m256 temp = _mm256_set1_ps(1.0)', + '__m256 temp = _mm256_set1_ps(1.0); temp = \ + _mm256_fmadd_ps(temp, temp, temp)', 'immintrin.h'), ('__attribute__((target("avx512f")))', 'attribute_target_avx512f_with_intrinsics', @@ -416,3 +417,41 @@ def long_double_representation(lines): else: # We never detected the after_sequence raise ValueError("Could not lock sequences (%s)" % saw) + + +def check_for_right_shift_internal_compiler_error(cmd): + """ + On our arm CI, this fails with an internal compilation error + + The failure looks like the following, and can be reproduced on ARM64 GCC 5.4: + + <source>: In function 'right_shift': + <source>:4:20: internal compiler error: in expand_shift_1, at expmed.c:2349 + ip1[i] = ip1[i] >> in2; + ^ + Please submit a full bug report, + with preprocessed source if appropriate. + See <http://gcc.gnu.org/bugs.html> for instructions. + Compiler returned: 1 + + This function returns True if this compiler bug is present, and we need to + turn off optimization for the function + """ + cmd._check_compiler() + has_optimize = cmd.try_compile(textwrap.dedent("""\ + __attribute__((optimize("O3"))) void right_shift() {} + """), None, None) + if not has_optimize: + return False + + no_err = cmd.try_compile(textwrap.dedent("""\ + typedef long the_type; /* fails also for unsigned and long long */ + __attribute__((optimize("O3"))) void right_shift(the_type in2, the_type *ip1, int n) { + for (int i = 0; i < n; i++) { + if (in2 < (the_type)sizeof(the_type) * 8) { + ip1[i] = ip1[i] >> in2; + } + } + } + """), None, None) + return not no_err diff --git a/numpy/core/shape_base.py b/numpy/core/shape_base.py index 710f64827..31b1c20b9 100644 --- a/numpy/core/shape_base.py +++ b/numpy/core/shape_base.py @@ -9,8 +9,9 @@ import warnings from . import numeric as _nx from . import overrides -from .numeric import array, asanyarray, newaxis +from ._asarray import array, asanyarray from .multiarray import normalize_axis_index +from . import fromnumeric as _from_nx array_function_dispatch = functools.partial( @@ -123,7 +124,7 @@ def atleast_2d(*arys): if ary.ndim == 0: result = ary.reshape(1, 1) elif ary.ndim == 1: - result = ary[newaxis, :] + result = ary[_nx.newaxis, :] else: result = ary res.append(result) @@ -193,9 +194,9 @@ def atleast_3d(*arys): if ary.ndim == 0: result = ary.reshape(1, 1, 1) elif ary.ndim == 1: - result = ary[newaxis, :, newaxis] + result = ary[_nx.newaxis, :, _nx.newaxis] elif ary.ndim == 2: - result = ary[:, :, newaxis] + result = ary[:, :, _nx.newaxis] else: result = ary res.append(result) @@ -435,9 +436,9 @@ def stack(arrays, axis=0, out=None): # Internal functions to eliminate the overhead of repeated dispatch in one of # the two possible paths inside np.block. # Use getattr to protect against __array_function__ being disabled. -_size = getattr(_nx.size, '__wrapped__', _nx.size) -_ndim = getattr(_nx.ndim, '__wrapped__', _nx.ndim) -_concatenate = getattr(_nx.concatenate, '__wrapped__', _nx.concatenate) +_size = getattr(_from_nx.size, '__wrapped__', _from_nx.size) +_ndim = getattr(_from_nx.ndim, '__wrapped__', _from_nx.ndim) +_concatenate = getattr(_from_nx.concatenate, '__wrapped__', _from_nx.concatenate) def _block_format_index(index): @@ -471,7 +472,7 @@ def _block_check_depths_match(arrays, parent_index=[]): first_index : list of int The full index of an element from the bottom of the nesting in `arrays`. If any element at the bottom is an empty list, this will - refer to it, and the last index along the empty axis will be `None`. + refer to it, and the last index along the empty axis will be None. max_arr_ndim : int The maximum of the ndims of the arrays nested in `arrays`. final_size: int @@ -574,7 +575,7 @@ def _concatenate_shapes(shapes, axis): that was computed deeper in the recursion. These are returned as tuples to ensure that they can quickly be added - to existing slice tuple without creating a new tuple everytime. + to existing slice tuple without creating a new tuple every time. """ # Cache a result that will be reused. diff --git a/numpy/core/src/common/binop_override.h b/numpy/core/src/common/binop_override.h index 47df63e38..c5e7ab808 100644 --- a/numpy/core/src/common/binop_override.h +++ b/numpy/core/src/common/binop_override.h @@ -129,11 +129,14 @@ binop_should_defer(PyObject *self, PyObject *other, int inplace) * check whether __array_ufunc__ equals None. */ attr = PyArray_LookupSpecial(other, "__array_ufunc__"); - if (attr) { + if (attr != NULL) { defer = !inplace && (attr == Py_None); Py_DECREF(attr); return defer; } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } /* * Otherwise, we need to check for the legacy __array_priority__. But if * other.__class__ is a subtype of self.__class__, then it's already had diff --git a/numpy/core/src/common/cblasfuncs.c b/numpy/core/src/common/cblasfuncs.c index 39572fed4..e78587de0 100644 --- a/numpy/core/src/common/cblasfuncs.c +++ b/numpy/core/src/common/cblasfuncs.c @@ -24,28 +24,28 @@ static const float oneF[2] = {1.0, 0.0}, zeroF[2] = {0.0, 0.0}; static void gemm(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE transA, enum CBLAS_TRANSPOSE transB, - int m, int n, int k, - PyArrayObject *A, int lda, PyArrayObject *B, int ldb, PyArrayObject *R) + npy_intp m, npy_intp n, npy_intp k, + PyArrayObject *A, npy_intp lda, PyArrayObject *B, npy_intp ldb, PyArrayObject *R) { const void *Adata = PyArray_DATA(A), *Bdata = PyArray_DATA(B); void *Rdata = PyArray_DATA(R); - int ldc = PyArray_DIM(R, 1) > 1 ? PyArray_DIM(R, 1) : 1; + npy_intp ldc = PyArray_DIM(R, 1) > 1 ? PyArray_DIM(R, 1) : 1; switch (typenum) { case NPY_DOUBLE: - cblas_dgemm(order, transA, transB, m, n, k, 1., + CBLAS_FUNC(cblas_dgemm)(order, transA, transB, m, n, k, 1., Adata, lda, Bdata, ldb, 0., Rdata, ldc); break; case NPY_FLOAT: - cblas_sgemm(order, transA, transB, m, n, k, 1.f, + CBLAS_FUNC(cblas_sgemm)(order, transA, transB, m, n, k, 1.f, Adata, lda, Bdata, ldb, 0.f, Rdata, ldc); break; case NPY_CDOUBLE: - cblas_zgemm(order, transA, transB, m, n, k, oneD, + CBLAS_FUNC(cblas_zgemm)(order, transA, transB, m, n, k, oneD, Adata, lda, Bdata, ldb, zeroD, Rdata, ldc); break; case NPY_CFLOAT: - cblas_cgemm(order, transA, transB, m, n, k, oneF, + CBLAS_FUNC(cblas_cgemm)(order, transA, transB, m, n, k, oneF, Adata, lda, Bdata, ldb, zeroF, Rdata, ldc); break; } @@ -57,29 +57,29 @@ gemm(int typenum, enum CBLAS_ORDER order, */ static void gemv(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, - PyArrayObject *A, int lda, PyArrayObject *X, int incX, + PyArrayObject *A, npy_intp lda, PyArrayObject *X, npy_intp incX, PyArrayObject *R) { const void *Adata = PyArray_DATA(A), *Xdata = PyArray_DATA(X); void *Rdata = PyArray_DATA(R); - int m = PyArray_DIM(A, 0), n = PyArray_DIM(A, 1); + npy_intp m = PyArray_DIM(A, 0), n = PyArray_DIM(A, 1); switch (typenum) { case NPY_DOUBLE: - cblas_dgemv(order, trans, m, n, 1., Adata, lda, Xdata, incX, + CBLAS_FUNC(cblas_dgemv)(order, trans, m, n, 1., Adata, lda, Xdata, incX, 0., Rdata, 1); break; case NPY_FLOAT: - cblas_sgemv(order, trans, m, n, 1.f, Adata, lda, Xdata, incX, + CBLAS_FUNC(cblas_sgemv)(order, trans, m, n, 1.f, Adata, lda, Xdata, incX, 0.f, Rdata, 1); break; case NPY_CDOUBLE: - cblas_zgemv(order, trans, m, n, oneD, Adata, lda, Xdata, incX, + CBLAS_FUNC(cblas_zgemv)(order, trans, m, n, oneD, Adata, lda, Xdata, incX, zeroD, Rdata, 1); break; case NPY_CFLOAT: - cblas_cgemv(order, trans, m, n, oneF, Adata, lda, Xdata, incX, + CBLAS_FUNC(cblas_cgemv)(order, trans, m, n, oneF, Adata, lda, Xdata, incX, zeroF, Rdata, 1); break; } @@ -91,19 +91,19 @@ gemv(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, */ static void syrk(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, - int n, int k, - PyArrayObject *A, int lda, PyArrayObject *R) + npy_intp n, npy_intp k, + PyArrayObject *A, npy_intp lda, PyArrayObject *R) { const void *Adata = PyArray_DATA(A); void *Rdata = PyArray_DATA(R); - int ldc = PyArray_DIM(R, 1) > 1 ? PyArray_DIM(R, 1) : 1; + npy_intp ldc = PyArray_DIM(R, 1) > 1 ? PyArray_DIM(R, 1) : 1; npy_intp i; npy_intp j; switch (typenum) { case NPY_DOUBLE: - cblas_dsyrk(order, CblasUpper, trans, n, k, 1., + CBLAS_FUNC(cblas_dsyrk)(order, CblasUpper, trans, n, k, 1., Adata, lda, 0., Rdata, ldc); for (i = 0; i < n; i++) { @@ -114,7 +114,7 @@ syrk(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, } break; case NPY_FLOAT: - cblas_ssyrk(order, CblasUpper, trans, n, k, 1.f, + CBLAS_FUNC(cblas_ssyrk)(order, CblasUpper, trans, n, k, 1.f, Adata, lda, 0.f, Rdata, ldc); for (i = 0; i < n; i++) { @@ -125,7 +125,7 @@ syrk(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, } break; case NPY_CDOUBLE: - cblas_zsyrk(order, CblasUpper, trans, n, k, oneD, + CBLAS_FUNC(cblas_zsyrk)(order, CblasUpper, trans, n, k, oneD, Adata, lda, zeroD, Rdata, ldc); for (i = 0; i < n; i++) { @@ -136,7 +136,7 @@ syrk(int typenum, enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE trans, } break; case NPY_CFLOAT: - cblas_csyrk(order, CblasUpper, trans, n, k, oneF, + CBLAS_FUNC(cblas_csyrk)(order, CblasUpper, trans, n, k, oneF, Adata, lda, zeroF, Rdata, ldc); for (i = 0; i < n; i++) { @@ -222,7 +222,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, PyArrayObject *out) { PyArrayObject *result = NULL, *out_buf = NULL; - int j, lda, ldb; + npy_intp j, lda, ldb; npy_intp l; int nd; npy_intp ap1stride = 0; @@ -385,14 +385,15 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, *((double *)PyArray_DATA(ap1)); } else if (ap1shape != _matrix) { - cblas_daxpy(l, + CBLAS_FUNC(cblas_daxpy)(l, *((double *)PyArray_DATA(ap2)), (double *)PyArray_DATA(ap1), ap1stride/sizeof(double), (double *)PyArray_DATA(out_buf), 1); } else { - int maxind, oind, i, a1s, outs; + int maxind, oind; + npy_intp i, a1s, outs; char *ptr, *optr; double val; @@ -405,7 +406,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, a1s = PyArray_STRIDE(ap1, maxind) / sizeof(double); outs = PyArray_STRIDE(out_buf, maxind) / sizeof(double); for (i = 0; i < PyArray_DIM(ap1, oind); i++) { - cblas_daxpy(l, val, (double *)ptr, a1s, + CBLAS_FUNC(cblas_daxpy)(l, val, (double *)ptr, a1s, (double *)optr, outs); ptr += PyArray_STRIDE(ap1, oind); optr += PyArray_STRIDE(out_buf, oind); @@ -423,14 +424,15 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, res->imag = ptr1->real * ptr2->imag + ptr1->imag * ptr2->real; } else if (ap1shape != _matrix) { - cblas_zaxpy(l, + CBLAS_FUNC(cblas_zaxpy)(l, (double *)PyArray_DATA(ap2), (double *)PyArray_DATA(ap1), ap1stride/sizeof(npy_cdouble), (double *)PyArray_DATA(out_buf), 1); } else { - int maxind, oind, i, a1s, outs; + int maxind, oind; + npy_intp i, a1s, outs; char *ptr, *optr; double *pval; @@ -443,7 +445,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, a1s = PyArray_STRIDE(ap1, maxind) / sizeof(npy_cdouble); outs = PyArray_STRIDE(out_buf, maxind) / sizeof(npy_cdouble); for (i = 0; i < PyArray_DIM(ap1, oind); i++) { - cblas_zaxpy(l, pval, (double *)ptr, a1s, + CBLAS_FUNC(cblas_zaxpy)(l, pval, (double *)ptr, a1s, (double *)optr, outs); ptr += PyArray_STRIDE(ap1, oind); optr += PyArray_STRIDE(out_buf, oind); @@ -456,14 +458,15 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, *((float *)PyArray_DATA(ap1)); } else if (ap1shape != _matrix) { - cblas_saxpy(l, + CBLAS_FUNC(cblas_saxpy)(l, *((float *)PyArray_DATA(ap2)), (float *)PyArray_DATA(ap1), ap1stride/sizeof(float), (float *)PyArray_DATA(out_buf), 1); } else { - int maxind, oind, i, a1s, outs; + int maxind, oind; + npy_intp i, a1s, outs; char *ptr, *optr; float val; @@ -476,7 +479,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, a1s = PyArray_STRIDE(ap1, maxind) / sizeof(float); outs = PyArray_STRIDE(out_buf, maxind) / sizeof(float); for (i = 0; i < PyArray_DIM(ap1, oind); i++) { - cblas_saxpy(l, val, (float *)ptr, a1s, + CBLAS_FUNC(cblas_saxpy)(l, val, (float *)ptr, a1s, (float *)optr, outs); ptr += PyArray_STRIDE(ap1, oind); optr += PyArray_STRIDE(out_buf, oind); @@ -494,14 +497,15 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, res->imag = ptr1->real * ptr2->imag + ptr1->imag * ptr2->real; } else if (ap1shape != _matrix) { - cblas_caxpy(l, + CBLAS_FUNC(cblas_caxpy)(l, (float *)PyArray_DATA(ap2), (float *)PyArray_DATA(ap1), ap1stride/sizeof(npy_cfloat), (float *)PyArray_DATA(out_buf), 1); } else { - int maxind, oind, i, a1s, outs; + int maxind, oind; + npy_intp i, a1s, outs; char *ptr, *optr; float *pval; @@ -514,7 +518,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, a1s = PyArray_STRIDE(ap1, maxind) / sizeof(npy_cfloat); outs = PyArray_STRIDE(out_buf, maxind) / sizeof(npy_cfloat); for (i = 0; i < PyArray_DIM(ap1, oind); i++) { - cblas_caxpy(l, pval, (float *)ptr, a1s, + CBLAS_FUNC(cblas_caxpy)(l, pval, (float *)ptr, a1s, (float *)optr, outs); ptr += PyArray_STRIDE(ap1, oind); optr += PyArray_STRIDE(out_buf, oind); @@ -537,7 +541,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, /* Matrix vector multiplication -- Level 2 BLAS */ /* lda must be MAX(M,1) */ enum CBLAS_ORDER Order; - int ap2s; + npy_intp ap2s; if (!PyArray_ISONESEGMENT(ap1)) { PyObject *new; @@ -564,7 +568,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, else if (ap1shape != _matrix && ap2shape == _matrix) { /* Vector matrix multiplication -- Level 2 BLAS */ enum CBLAS_ORDER Order; - int ap1s; + npy_intp ap1s; if (!PyArray_ISONESEGMENT(ap2)) { PyObject *new; @@ -601,7 +605,7 @@ cblas_matrixproduct(int typenum, PyArrayObject *ap1, PyArrayObject *ap2, */ enum CBLAS_ORDER Order; enum CBLAS_TRANSPOSE Trans1, Trans2; - int M, N, L; + npy_intp M, N, L; /* Optimization possible: */ /* diff --git a/numpy/core/src/common/get_attr_string.h b/numpy/core/src/common/get_attr_string.h index d458d9550..d3401aea6 100644 --- a/numpy/core/src/common/get_attr_string.h +++ b/numpy/core/src/common/get_attr_string.h @@ -40,18 +40,14 @@ _is_basic_python_type(PyTypeObject *tp) } /* - * Stripped down version of PyObject_GetAttrString, - * avoids lookups for None, tuple, and List objects, - * and doesn't create a PyErr since this code ignores it. + * Stripped down version of PyObject_GetAttrString(obj, name) that does not + * raise PyExc_AttributeError. * - * This can be much faster then PyObject_GetAttrString where - * exceptions are not used by caller. + * This allows it to avoid creating then discarding exception objects when + * performing lookups on objects without any attributes. * - * 'obj' is the object to search for attribute. - * - * 'name' is the attribute to search for. - * - * Returns attribute value on success, NULL on failure. + * Returns attribute value on success, NULL without an exception set if + * there is no such attribute, and NULL with an exception on failure. */ static NPY_INLINE PyObject * maybe_get_attr(PyObject *obj, char *name) @@ -62,7 +58,7 @@ maybe_get_attr(PyObject *obj, char *name) /* Attribute referenced by (char *)name */ if (tp->tp_getattr != NULL) { res = (*tp->tp_getattr)(obj, name); - if (res == NULL) { + if (res == NULL && PyErr_ExceptionMatches(PyExc_AttributeError)) { PyErr_Clear(); } } @@ -78,7 +74,7 @@ maybe_get_attr(PyObject *obj, char *name) } res = (*tp->tp_getattro)(obj, w); Py_DECREF(w); - if (res == NULL) { + if (res == NULL && PyErr_ExceptionMatches(PyExc_AttributeError)) { PyErr_Clear(); } } diff --git a/numpy/core/src/common/npy_cblas.h b/numpy/core/src/common/npy_cblas.h index a083f3bcc..97308238a 100644 --- a/numpy/core/src/common/npy_cblas.h +++ b/numpy/core/src/common/npy_cblas.h @@ -17,565 +17,47 @@ extern "C" /* * Enumerated and derived types */ -#define CBLAS_INDEX size_t /* this may vary between platforms */ - enum CBLAS_ORDER {CblasRowMajor=101, CblasColMajor=102}; enum CBLAS_TRANSPOSE {CblasNoTrans=111, CblasTrans=112, CblasConjTrans=113}; enum CBLAS_UPLO {CblasUpper=121, CblasLower=122}; enum CBLAS_DIAG {CblasNonUnit=131, CblasUnit=132}; enum CBLAS_SIDE {CblasLeft=141, CblasRight=142}; -/* - * =========================================================================== - * Prototypes for level 1 BLAS functions (complex are recast as routines) - * =========================================================================== - */ -float cblas_sdsdot(const int N, const float alpha, const float *X, - const int incX, const float *Y, const int incY); -double cblas_dsdot(const int N, const float *X, const int incX, const float *Y, - const int incY); -float cblas_sdot(const int N, const float *X, const int incX, - const float *Y, const int incY); -double cblas_ddot(const int N, const double *X, const int incX, - const double *Y, const int incY); - -/* - * Functions having prefixes Z and C only - */ -void cblas_cdotu_sub(const int N, const void *X, const int incX, - const void *Y, const int incY, void *dotu); -void cblas_cdotc_sub(const int N, const void *X, const int incX, - const void *Y, const int incY, void *dotc); - -void cblas_zdotu_sub(const int N, const void *X, const int incX, - const void *Y, const int incY, void *dotu); -void cblas_zdotc_sub(const int N, const void *X, const int incX, - const void *Y, const int incY, void *dotc); - - -/* - * Functions having prefixes S D SC DZ - */ -float cblas_snrm2(const int N, const float *X, const int incX); -float cblas_sasum(const int N, const float *X, const int incX); - -double cblas_dnrm2(const int N, const double *X, const int incX); -double cblas_dasum(const int N, const double *X, const int incX); - -float cblas_scnrm2(const int N, const void *X, const int incX); -float cblas_scasum(const int N, const void *X, const int incX); - -double cblas_dznrm2(const int N, const void *X, const int incX); -double cblas_dzasum(const int N, const void *X, const int incX); - - -/* - * Functions having standard 4 prefixes (S D C Z) - */ -CBLAS_INDEX cblas_isamax(const int N, const float *X, const int incX); -CBLAS_INDEX cblas_idamax(const int N, const double *X, const int incX); -CBLAS_INDEX cblas_icamax(const int N, const void *X, const int incX); -CBLAS_INDEX cblas_izamax(const int N, const void *X, const int incX); - -/* - * =========================================================================== - * Prototypes for level 1 BLAS routines - * =========================================================================== - */ - -/* - * Routines with standard 4 prefixes (s, d, c, z) - */ -void cblas_sswap(const int N, float *X, const int incX, - float *Y, const int incY); -void cblas_scopy(const int N, const float *X, const int incX, - float *Y, const int incY); -void cblas_saxpy(const int N, const float alpha, const float *X, - const int incX, float *Y, const int incY); - -void cblas_dswap(const int N, double *X, const int incX, - double *Y, const int incY); -void cblas_dcopy(const int N, const double *X, const int incX, - double *Y, const int incY); -void cblas_daxpy(const int N, const double alpha, const double *X, - const int incX, double *Y, const int incY); - -void cblas_cswap(const int N, void *X, const int incX, - void *Y, const int incY); -void cblas_ccopy(const int N, const void *X, const int incX, - void *Y, const int incY); -void cblas_caxpy(const int N, const void *alpha, const void *X, - const int incX, void *Y, const int incY); - -void cblas_zswap(const int N, void *X, const int incX, - void *Y, const int incY); -void cblas_zcopy(const int N, const void *X, const int incX, - void *Y, const int incY); -void cblas_zaxpy(const int N, const void *alpha, const void *X, - const int incX, void *Y, const int incY); - - -/* - * Routines with S and D prefix only - */ -void cblas_srotg(float *a, float *b, float *c, float *s); -void cblas_srotmg(float *d1, float *d2, float *b1, const float b2, float *P); -void cblas_srot(const int N, float *X, const int incX, - float *Y, const int incY, const float c, const float s); -void cblas_srotm(const int N, float *X, const int incX, - float *Y, const int incY, const float *P); - -void cblas_drotg(double *a, double *b, double *c, double *s); -void cblas_drotmg(double *d1, double *d2, double *b1, const double b2, double *P); -void cblas_drot(const int N, double *X, const int incX, - double *Y, const int incY, const double c, const double s); -void cblas_drotm(const int N, double *X, const int incX, - double *Y, const int incY, const double *P); - - -/* - * Routines with S D C Z CS and ZD prefixes - */ -void cblas_sscal(const int N, const float alpha, float *X, const int incX); -void cblas_dscal(const int N, const double alpha, double *X, const int incX); -void cblas_cscal(const int N, const void *alpha, void *X, const int incX); -void cblas_zscal(const int N, const void *alpha, void *X, const int incX); -void cblas_csscal(const int N, const float alpha, void *X, const int incX); -void cblas_zdscal(const int N, const double alpha, void *X, const int incX); - -/* - * =========================================================================== - * Prototypes for level 2 BLAS - * =========================================================================== - */ - -/* - * Routines with standard 4 prefixes (S, D, C, Z) - */ -void cblas_sgemv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const float alpha, const float *A, const int lda, - const float *X, const int incX, const float beta, - float *Y, const int incY); -void cblas_sgbmv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const int KL, const int KU, const float alpha, - const float *A, const int lda, const float *X, - const int incX, const float beta, float *Y, const int incY); -void cblas_strmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const float *A, const int lda, - float *X, const int incX); -void cblas_stbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const float *A, const int lda, - float *X, const int incX); -void cblas_stpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const float *Ap, float *X, const int incX); -void cblas_strsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const float *A, const int lda, float *X, - const int incX); -void cblas_stbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const float *A, const int lda, - float *X, const int incX); -void cblas_stpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const float *Ap, float *X, const int incX); - -void cblas_dgemv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const double alpha, const double *A, const int lda, - const double *X, const int incX, const double beta, - double *Y, const int incY); -void cblas_dgbmv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const int KL, const int KU, const double alpha, - const double *A, const int lda, const double *X, - const int incX, const double beta, double *Y, const int incY); -void cblas_dtrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const double *A, const int lda, - double *X, const int incX); -void cblas_dtbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const double *A, const int lda, - double *X, const int incX); -void cblas_dtpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const double *Ap, double *X, const int incX); -void cblas_dtrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const double *A, const int lda, double *X, - const int incX); -void cblas_dtbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const double *A, const int lda, - double *X, const int incX); -void cblas_dtpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const double *Ap, double *X, const int incX); - -void cblas_cgemv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *X, const int incX, const void *beta, - void *Y, const int incY); -void cblas_cgbmv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const int KL, const int KU, const void *alpha, - const void *A, const int lda, const void *X, - const int incX, const void *beta, void *Y, const int incY); -void cblas_ctrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *A, const int lda, - void *X, const int incX); -void cblas_ctbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const void *A, const int lda, - void *X, const int incX); -void cblas_ctpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *Ap, void *X, const int incX); -void cblas_ctrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *A, const int lda, void *X, - const int incX); -void cblas_ctbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const void *A, const int lda, - void *X, const int incX); -void cblas_ctpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *Ap, void *X, const int incX); - -void cblas_zgemv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *X, const int incX, const void *beta, - void *Y, const int incY); -void cblas_zgbmv(const enum CBLAS_ORDER order, - const enum CBLAS_TRANSPOSE TransA, const int M, const int N, - const int KL, const int KU, const void *alpha, - const void *A, const int lda, const void *X, - const int incX, const void *beta, void *Y, const int incY); -void cblas_ztrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *A, const int lda, - void *X, const int incX); -void cblas_ztbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const void *A, const int lda, - void *X, const int incX); -void cblas_ztpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *Ap, void *X, const int incX); -void cblas_ztrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *A, const int lda, void *X, - const int incX); -void cblas_ztbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const int K, const void *A, const int lda, - void *X, const int incX); -void cblas_ztpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, - const int N, const void *Ap, void *X, const int incX); - - -/* - * Routines with S and D prefixes only - */ -void cblas_ssymv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *A, - const int lda, const float *X, const int incX, - const float beta, float *Y, const int incY); -void cblas_ssbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const int K, const float alpha, const float *A, - const int lda, const float *X, const int incX, - const float beta, float *Y, const int incY); -void cblas_sspmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *Ap, - const float *X, const int incX, - const float beta, float *Y, const int incY); -void cblas_sger(const enum CBLAS_ORDER order, const int M, const int N, - const float alpha, const float *X, const int incX, - const float *Y, const int incY, float *A, const int lda); -void cblas_ssyr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *X, - const int incX, float *A, const int lda); -void cblas_sspr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *X, - const int incX, float *Ap); -void cblas_ssyr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *X, - const int incX, const float *Y, const int incY, float *A, - const int lda); -void cblas_sspr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const float *X, - const int incX, const float *Y, const int incY, float *A); - -void cblas_dsymv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *A, - const int lda, const double *X, const int incX, - const double beta, double *Y, const int incY); -void cblas_dsbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const int K, const double alpha, const double *A, - const int lda, const double *X, const int incX, - const double beta, double *Y, const int incY); -void cblas_dspmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *Ap, - const double *X, const int incX, - const double beta, double *Y, const int incY); -void cblas_dger(const enum CBLAS_ORDER order, const int M, const int N, - const double alpha, const double *X, const int incX, - const double *Y, const int incY, double *A, const int lda); -void cblas_dsyr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *X, - const int incX, double *A, const int lda); -void cblas_dspr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *X, - const int incX, double *Ap); -void cblas_dsyr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *X, - const int incX, const double *Y, const int incY, double *A, - const int lda); -void cblas_dspr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const double *X, - const int incX, const double *Y, const int incY, double *A); - - -/* - * Routines with C and Z prefixes only - */ -void cblas_chemv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const void *alpha, const void *A, - const int lda, const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_chbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const int K, const void *alpha, const void *A, - const int lda, const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_chpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const void *alpha, const void *Ap, - const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_cgeru(const enum CBLAS_ORDER order, const int M, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_cgerc(const enum CBLAS_ORDER order, const int M, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_cher(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const void *X, const int incX, - void *A, const int lda); -void cblas_chpr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const float alpha, const void *X, - const int incX, void *A); -void cblas_cher2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_chpr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *Ap); - -void cblas_zhemv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const void *alpha, const void *A, - const int lda, const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_zhbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const int K, const void *alpha, const void *A, - const int lda, const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_zhpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const void *alpha, const void *Ap, - const void *X, const int incX, - const void *beta, void *Y, const int incY); -void cblas_zgeru(const enum CBLAS_ORDER order, const int M, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_zgerc(const enum CBLAS_ORDER order, const int M, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_zher(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const void *X, const int incX, - void *A, const int lda); -void cblas_zhpr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, - const int N, const double alpha, const void *X, - const int incX, void *A); -void cblas_zher2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *A, const int lda); -void cblas_zhpr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const int N, - const void *alpha, const void *X, const int incX, - const void *Y, const int incY, void *Ap); +#define CBLAS_INDEX size_t /* this may vary between platforms */ -/* - * =========================================================================== - * Prototypes for level 3 BLAS - * =========================================================================== - */ +#ifdef NO_APPEND_FORTRAN +#define BLAS_FORTRAN_SUFFIX +#else +#define BLAS_FORTRAN_SUFFIX _ +#endif -/* - * Routines with standard 4 prefixes (S, D, C, Z) - */ -void cblas_sgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_TRANSPOSE TransB, const int M, const int N, - const int K, const float alpha, const float *A, - const int lda, const float *B, const int ldb, - const float beta, float *C, const int ldc); -void cblas_ssymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const float alpha, const float *A, const int lda, - const float *B, const int ldb, const float beta, - float *C, const int ldc); -void cblas_ssyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const float alpha, const float *A, const int lda, - const float beta, float *C, const int ldc); -void cblas_ssyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const float alpha, const float *A, const int lda, - const float *B, const int ldb, const float beta, - float *C, const int ldc); -void cblas_strmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const float alpha, const float *A, const int lda, - float *B, const int ldb); -void cblas_strsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const float alpha, const float *A, const int lda, - float *B, const int ldb); +#ifndef BLAS_SYMBOL_PREFIX +#define BLAS_SYMBOL_PREFIX +#endif -void cblas_dgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_TRANSPOSE TransB, const int M, const int N, - const int K, const double alpha, const double *A, - const int lda, const double *B, const int ldb, - const double beta, double *C, const int ldc); -void cblas_dsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const double alpha, const double *A, const int lda, - const double *B, const int ldb, const double beta, - double *C, const int ldc); -void cblas_dsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const double alpha, const double *A, const int lda, - const double beta, double *C, const int ldc); -void cblas_dsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const double alpha, const double *A, const int lda, - const double *B, const int ldb, const double beta, - double *C, const int ldc); -void cblas_dtrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const double alpha, const double *A, const int lda, - double *B, const int ldb); -void cblas_dtrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const double alpha, const double *A, const int lda, - double *B, const int ldb); +#ifndef BLAS_SYMBOL_SUFFIX +#define BLAS_SYMBOL_SUFFIX +#endif -void cblas_cgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_TRANSPOSE TransB, const int M, const int N, - const int K, const void *alpha, const void *A, - const int lda, const void *B, const int ldb, - const void *beta, void *C, const int ldc); -void cblas_csymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_csyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *beta, void *C, const int ldc); -void cblas_csyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_ctrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const void *alpha, const void *A, const int lda, - void *B, const int ldb); -void cblas_ctrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const void *alpha, const void *A, const int lda, - void *B, const int ldb); +#define BLAS_FUNC_CONCAT(name,prefix,suffix,suffix2) prefix ## name ## suffix ## suffix2 +#define BLAS_FUNC_EXPAND(name,prefix,suffix,suffix2) BLAS_FUNC_CONCAT(name,prefix,suffix,suffix2) -void cblas_zgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_TRANSPOSE TransB, const int M, const int N, - const int K, const void *alpha, const void *A, - const int lda, const void *B, const int ldb, - const void *beta, void *C, const int ldc); -void cblas_zsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_zsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *beta, void *C, const int ldc); -void cblas_zsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_ztrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const void *alpha, const void *A, const int lda, - void *B, const int ldb); -void cblas_ztrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, - const enum CBLAS_DIAG Diag, const int M, const int N, - const void *alpha, const void *A, const int lda, - void *B, const int ldb); +#define CBLAS_FUNC(name) BLAS_FUNC_EXPAND(name,BLAS_SYMBOL_PREFIX,,BLAS_SYMBOL_SUFFIX) +#define BLAS_FUNC(name) BLAS_FUNC_EXPAND(name,BLAS_SYMBOL_PREFIX,BLAS_FORTRAN_SUFFIX,BLAS_SYMBOL_SUFFIX) +#ifdef HAVE_BLAS_ILP64 +#define CBLAS_INT npy_int64 +#else +#define CBLAS_INT int +#endif -/* - * Routines with prefixes C and Z only - */ -void cblas_chemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_cherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const float alpha, const void *A, const int lda, - const float beta, void *C, const int ldc); -void cblas_cher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const float beta, - void *C, const int ldc); +#define BLASNAME(name) CBLAS_FUNC(name) +#define BLASINT CBLAS_INT -void cblas_zhemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, - const enum CBLAS_UPLO Uplo, const int M, const int N, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const void *beta, - void *C, const int ldc); -void cblas_zherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const double alpha, const void *A, const int lda, - const double beta, void *C, const int ldc); -void cblas_zher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, - const enum CBLAS_TRANSPOSE Trans, const int N, const int K, - const void *alpha, const void *A, const int lda, - const void *B, const int ldb, const double beta, - void *C, const int ldc); +#include "npy_cblas_base.h" -void cblas_xerbla(int p, const char *rout, const char *form, ...); +#undef BLASINT +#undef BLASNAME #ifdef __cplusplus } diff --git a/numpy/core/src/common/npy_cblas_base.h b/numpy/core/src/common/npy_cblas_base.h new file mode 100644 index 000000000..792b6f09e --- /dev/null +++ b/numpy/core/src/common/npy_cblas_base.h @@ -0,0 +1,557 @@ +/* + * This header provides numpy a consistent interface to CBLAS code. It is needed + * because not all providers of cblas provide cblas.h. For instance, MKL provides + * mkl_cblas.h and also typedefs the CBLAS_XXX enums. + */ + +/* + * =========================================================================== + * Prototypes for level 1 BLAS functions (complex are recast as routines) + * =========================================================================== + */ +float BLASNAME(cblas_sdsdot)(const BLASINT N, const float alpha, const float *X, + const BLASINT incX, const float *Y, const BLASINT incY); +double BLASNAME(cblas_dsdot)(const BLASINT N, const float *X, const BLASINT incX, const float *Y, + const BLASINT incY); +float BLASNAME(cblas_sdot)(const BLASINT N, const float *X, const BLASINT incX, + const float *Y, const BLASINT incY); +double BLASNAME(cblas_ddot)(const BLASINT N, const double *X, const BLASINT incX, + const double *Y, const BLASINT incY); + +/* + * Functions having prefixes Z and C only + */ +void BLASNAME(cblas_cdotu_sub)(const BLASINT N, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *dotu); +void BLASNAME(cblas_cdotc_sub)(const BLASINT N, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *dotc); + +void BLASNAME(cblas_zdotu_sub)(const BLASINT N, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *dotu); +void BLASNAME(cblas_zdotc_sub)(const BLASINT N, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *dotc); + + +/* + * Functions having prefixes S D SC DZ + */ +float BLASNAME(cblas_snrm2)(const BLASINT N, const float *X, const BLASINT incX); +float BLASNAME(cblas_sasum)(const BLASINT N, const float *X, const BLASINT incX); + +double BLASNAME(cblas_dnrm2)(const BLASINT N, const double *X, const BLASINT incX); +double BLASNAME(cblas_dasum)(const BLASINT N, const double *X, const BLASINT incX); + +float BLASNAME(cblas_scnrm2)(const BLASINT N, const void *X, const BLASINT incX); +float BLASNAME(cblas_scasum)(const BLASINT N, const void *X, const BLASINT incX); + +double BLASNAME(cblas_dznrm2)(const BLASINT N, const void *X, const BLASINT incX); +double BLASNAME(cblas_dzasum)(const BLASINT N, const void *X, const BLASINT incX); + + +/* + * Functions having standard 4 prefixes (S D C Z) + */ +CBLAS_INDEX BLASNAME(cblas_isamax)(const BLASINT N, const float *X, const BLASINT incX); +CBLAS_INDEX BLASNAME(cblas_idamax)(const BLASINT N, const double *X, const BLASINT incX); +CBLAS_INDEX BLASNAME(cblas_icamax)(const BLASINT N, const void *X, const BLASINT incX); +CBLAS_INDEX BLASNAME(cblas_izamax)(const BLASINT N, const void *X, const BLASINT incX); + +/* + * =========================================================================== + * Prototypes for level 1 BLAS routines + * =========================================================================== + */ + +/* + * Routines with standard 4 prefixes (s, d, c, z) + */ +void BLASNAME(cblas_sswap)(const BLASINT N, float *X, const BLASINT incX, + float *Y, const BLASINT incY); +void BLASNAME(cblas_scopy)(const BLASINT N, const float *X, const BLASINT incX, + float *Y, const BLASINT incY); +void BLASNAME(cblas_saxpy)(const BLASINT N, const float alpha, const float *X, + const BLASINT incX, float *Y, const BLASINT incY); + +void BLASNAME(cblas_dswap)(const BLASINT N, double *X, const BLASINT incX, + double *Y, const BLASINT incY); +void BLASNAME(cblas_dcopy)(const BLASINT N, const double *X, const BLASINT incX, + double *Y, const BLASINT incY); +void BLASNAME(cblas_daxpy)(const BLASINT N, const double alpha, const double *X, + const BLASINT incX, double *Y, const BLASINT incY); + +void BLASNAME(cblas_cswap)(const BLASINT N, void *X, const BLASINT incX, + void *Y, const BLASINT incY); +void BLASNAME(cblas_ccopy)(const BLASINT N, const void *X, const BLASINT incX, + void *Y, const BLASINT incY); +void BLASNAME(cblas_caxpy)(const BLASINT N, const void *alpha, const void *X, + const BLASINT incX, void *Y, const BLASINT incY); + +void BLASNAME(cblas_zswap)(const BLASINT N, void *X, const BLASINT incX, + void *Y, const BLASINT incY); +void BLASNAME(cblas_zcopy)(const BLASINT N, const void *X, const BLASINT incX, + void *Y, const BLASINT incY); +void BLASNAME(cblas_zaxpy)(const BLASINT N, const void *alpha, const void *X, + const BLASINT incX, void *Y, const BLASINT incY); + + +/* + * Routines with S and D prefix only + */ +void BLASNAME(cblas_srotg)(float *a, float *b, float *c, float *s); +void BLASNAME(cblas_srotmg)(float *d1, float *d2, float *b1, const float b2, float *P); +void BLASNAME(cblas_srot)(const BLASINT N, float *X, const BLASINT incX, + float *Y, const BLASINT incY, const float c, const float s); +void BLASNAME(cblas_srotm)(const BLASINT N, float *X, const BLASINT incX, + float *Y, const BLASINT incY, const float *P); + +void BLASNAME(cblas_drotg)(double *a, double *b, double *c, double *s); +void BLASNAME(cblas_drotmg)(double *d1, double *d2, double *b1, const double b2, double *P); +void BLASNAME(cblas_drot)(const BLASINT N, double *X, const BLASINT incX, + double *Y, const BLASINT incY, const double c, const double s); +void BLASNAME(cblas_drotm)(const BLASINT N, double *X, const BLASINT incX, + double *Y, const BLASINT incY, const double *P); + + +/* + * Routines with S D C Z CS and ZD prefixes + */ +void BLASNAME(cblas_sscal)(const BLASINT N, const float alpha, float *X, const BLASINT incX); +void BLASNAME(cblas_dscal)(const BLASINT N, const double alpha, double *X, const BLASINT incX); +void BLASNAME(cblas_cscal)(const BLASINT N, const void *alpha, void *X, const BLASINT incX); +void BLASNAME(cblas_zscal)(const BLASINT N, const void *alpha, void *X, const BLASINT incX); +void BLASNAME(cblas_csscal)(const BLASINT N, const float alpha, void *X, const BLASINT incX); +void BLASNAME(cblas_zdscal)(const BLASINT N, const double alpha, void *X, const BLASINT incX); + +/* + * =========================================================================== + * Prototypes for level 2 BLAS + * =========================================================================== + */ + +/* + * Routines with standard 4 prefixes (S, D, C, Z) + */ +void BLASNAME(cblas_sgemv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const float alpha, const float *A, const BLASINT lda, + const float *X, const BLASINT incX, const float beta, + float *Y, const BLASINT incY); +void BLASNAME(cblas_sgbmv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const BLASINT KL, const BLASINT KU, const float alpha, + const float *A, const BLASINT lda, const float *X, + const BLASINT incX, const float beta, float *Y, const BLASINT incY); +void BLASNAME(cblas_strmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const float *A, const BLASINT lda, + float *X, const BLASINT incX); +void BLASNAME(cblas_stbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const float *A, const BLASINT lda, + float *X, const BLASINT incX); +void BLASNAME(cblas_stpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const float *Ap, float *X, const BLASINT incX); +void BLASNAME(cblas_strsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const float *A, const BLASINT lda, float *X, + const BLASINT incX); +void BLASNAME(cblas_stbsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const float *A, const BLASINT lda, + float *X, const BLASINT incX); +void BLASNAME(cblas_stpsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const float *Ap, float *X, const BLASINT incX); + +void BLASNAME(cblas_dgemv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const double alpha, const double *A, const BLASINT lda, + const double *X, const BLASINT incX, const double beta, + double *Y, const BLASINT incY); +void BLASNAME(cblas_dgbmv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const BLASINT KL, const BLASINT KU, const double alpha, + const double *A, const BLASINT lda, const double *X, + const BLASINT incX, const double beta, double *Y, const BLASINT incY); +void BLASNAME(cblas_dtrmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const double *A, const BLASINT lda, + double *X, const BLASINT incX); +void BLASNAME(cblas_dtbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const double *A, const BLASINT lda, + double *X, const BLASINT incX); +void BLASNAME(cblas_dtpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const double *Ap, double *X, const BLASINT incX); +void BLASNAME(cblas_dtrsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const double *A, const BLASINT lda, double *X, + const BLASINT incX); +void BLASNAME(cblas_dtbsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const double *A, const BLASINT lda, + double *X, const BLASINT incX); +void BLASNAME(cblas_dtpsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const double *Ap, double *X, const BLASINT incX); + +void BLASNAME(cblas_cgemv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *X, const BLASINT incX, const void *beta, + void *Y, const BLASINT incY); +void BLASNAME(cblas_cgbmv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const BLASINT KL, const BLASINT KU, const void *alpha, + const void *A, const BLASINT lda, const void *X, + const BLASINT incX, const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_ctrmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ctbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ctpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *Ap, void *X, const BLASINT incX); +void BLASNAME(cblas_ctrsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *A, const BLASINT lda, void *X, + const BLASINT incX); +void BLASNAME(cblas_ctbsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ctpsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *Ap, void *X, const BLASINT incX); + +void BLASNAME(cblas_zgemv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *X, const BLASINT incX, const void *beta, + void *Y, const BLASINT incY); +void BLASNAME(cblas_zgbmv)(const enum CBLAS_ORDER order, + const enum CBLAS_TRANSPOSE TransA, const BLASINT M, const BLASINT N, + const BLASINT KL, const BLASINT KU, const void *alpha, + const void *A, const BLASINT lda, const void *X, + const BLASINT incX, const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_ztrmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ztbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ztpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *Ap, void *X, const BLASINT incX); +void BLASNAME(cblas_ztrsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *A, const BLASINT lda, void *X, + const BLASINT incX); +void BLASNAME(cblas_ztbsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const BLASINT K, const void *A, const BLASINT lda, + void *X, const BLASINT incX); +void BLASNAME(cblas_ztpsv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, + const BLASINT N, const void *Ap, void *X, const BLASINT incX); + + +/* + * Routines with S and D prefixes only + */ +void BLASNAME(cblas_ssymv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *A, + const BLASINT lda, const float *X, const BLASINT incX, + const float beta, float *Y, const BLASINT incY); +void BLASNAME(cblas_ssbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const BLASINT K, const float alpha, const float *A, + const BLASINT lda, const float *X, const BLASINT incX, + const float beta, float *Y, const BLASINT incY); +void BLASNAME(cblas_sspmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *Ap, + const float *X, const BLASINT incX, + const float beta, float *Y, const BLASINT incY); +void BLASNAME(cblas_sger)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const float alpha, const float *X, const BLASINT incX, + const float *Y, const BLASINT incY, float *A, const BLASINT lda); +void BLASNAME(cblas_ssyr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *X, + const BLASINT incX, float *A, const BLASINT lda); +void BLASNAME(cblas_sspr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *X, + const BLASINT incX, float *Ap); +void BLASNAME(cblas_ssyr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *X, + const BLASINT incX, const float *Y, const BLASINT incY, float *A, + const BLASINT lda); +void BLASNAME(cblas_sspr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const float *X, + const BLASINT incX, const float *Y, const BLASINT incY, float *A); + +void BLASNAME(cblas_dsymv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *A, + const BLASINT lda, const double *X, const BLASINT incX, + const double beta, double *Y, const BLASINT incY); +void BLASNAME(cblas_dsbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const BLASINT K, const double alpha, const double *A, + const BLASINT lda, const double *X, const BLASINT incX, + const double beta, double *Y, const BLASINT incY); +void BLASNAME(cblas_dspmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *Ap, + const double *X, const BLASINT incX, + const double beta, double *Y, const BLASINT incY); +void BLASNAME(cblas_dger)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const double alpha, const double *X, const BLASINT incX, + const double *Y, const BLASINT incY, double *A, const BLASINT lda); +void BLASNAME(cblas_dsyr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *X, + const BLASINT incX, double *A, const BLASINT lda); +void BLASNAME(cblas_dspr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *X, + const BLASINT incX, double *Ap); +void BLASNAME(cblas_dsyr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *X, + const BLASINT incX, const double *Y, const BLASINT incY, double *A, + const BLASINT lda); +void BLASNAME(cblas_dspr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const double *X, + const BLASINT incX, const double *Y, const BLASINT incY, double *A); + + +/* + * Routines with C and Z prefixes only + */ +void BLASNAME(cblas_chemv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const void *alpha, const void *A, + const BLASINT lda, const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_chbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const BLASINT K, const void *alpha, const void *A, + const BLASINT lda, const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_chpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const void *alpha, const void *Ap, + const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_cgeru)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_cgerc)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_cher)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const void *X, const BLASINT incX, + void *A, const BLASINT lda); +void BLASNAME(cblas_chpr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const float alpha, const void *X, + const BLASINT incX, void *A); +void BLASNAME(cblas_cher2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_chpr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *Ap); + +void BLASNAME(cblas_zhemv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const void *alpha, const void *A, + const BLASINT lda, const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_zhbmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const BLASINT K, const void *alpha, const void *A, + const BLASINT lda, const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_zhpmv)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const void *alpha, const void *Ap, + const void *X, const BLASINT incX, + const void *beta, void *Y, const BLASINT incY); +void BLASNAME(cblas_zgeru)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_zgerc)(const enum CBLAS_ORDER order, const BLASINT M, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_zher)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const void *X, const BLASINT incX, + void *A, const BLASINT lda); +void BLASNAME(cblas_zhpr)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, + const BLASINT N, const double alpha, const void *X, + const BLASINT incX, void *A); +void BLASNAME(cblas_zher2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *A, const BLASINT lda); +void BLASNAME(cblas_zhpr2)(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const BLASINT N, + const void *alpha, const void *X, const BLASINT incX, + const void *Y, const BLASINT incY, void *Ap); + +/* + * =========================================================================== + * Prototypes for level 3 BLAS + * =========================================================================== + */ + +/* + * Routines with standard 4 prefixes (S, D, C, Z) + */ +void BLASNAME(cblas_sgemm)(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_TRANSPOSE TransB, const BLASINT M, const BLASINT N, + const BLASINT K, const float alpha, const float *A, + const BLASINT lda, const float *B, const BLASINT ldb, + const float beta, float *C, const BLASINT ldc); +void BLASNAME(cblas_ssymm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const float alpha, const float *A, const BLASINT lda, + const float *B, const BLASINT ldb, const float beta, + float *C, const BLASINT ldc); +void BLASNAME(cblas_ssyrk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const float alpha, const float *A, const BLASINT lda, + const float beta, float *C, const BLASINT ldc); +void BLASNAME(cblas_ssyr2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const float alpha, const float *A, const BLASINT lda, + const float *B, const BLASINT ldb, const float beta, + float *C, const BLASINT ldc); +void BLASNAME(cblas_strmm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const float alpha, const float *A, const BLASINT lda, + float *B, const BLASINT ldb); +void BLASNAME(cblas_strsm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const float alpha, const float *A, const BLASINT lda, + float *B, const BLASINT ldb); + +void BLASNAME(cblas_dgemm)(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_TRANSPOSE TransB, const BLASINT M, const BLASINT N, + const BLASINT K, const double alpha, const double *A, + const BLASINT lda, const double *B, const BLASINT ldb, + const double beta, double *C, const BLASINT ldc); +void BLASNAME(cblas_dsymm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const double alpha, const double *A, const BLASINT lda, + const double *B, const BLASINT ldb, const double beta, + double *C, const BLASINT ldc); +void BLASNAME(cblas_dsyrk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const double alpha, const double *A, const BLASINT lda, + const double beta, double *C, const BLASINT ldc); +void BLASNAME(cblas_dsyr2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const double alpha, const double *A, const BLASINT lda, + const double *B, const BLASINT ldb, const double beta, + double *C, const BLASINT ldc); +void BLASNAME(cblas_dtrmm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const double alpha, const double *A, const BLASINT lda, + double *B, const BLASINT ldb); +void BLASNAME(cblas_dtrsm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const double alpha, const double *A, const BLASINT lda, + double *B, const BLASINT ldb); + +void BLASNAME(cblas_cgemm)(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_TRANSPOSE TransB, const BLASINT M, const BLASINT N, + const BLASINT K, const void *alpha, const void *A, + const BLASINT lda, const void *B, const BLASINT ldb, + const void *beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_csymm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_csyrk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_csyr2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_ctrmm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + void *B, const BLASINT ldb); +void BLASNAME(cblas_ctrsm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + void *B, const BLASINT ldb); + +void BLASNAME(cblas_zgemm)(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_TRANSPOSE TransB, const BLASINT M, const BLASINT N, + const BLASINT K, const void *alpha, const void *A, + const BLASINT lda, const void *B, const BLASINT ldb, + const void *beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_zsymm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_zsyrk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_zsyr2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_ztrmm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + void *B, const BLASINT ldb); +void BLASNAME(cblas_ztrsm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, + const enum CBLAS_DIAG Diag, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + void *B, const BLASINT ldb); + + +/* + * Routines with prefixes C and Z only + */ +void BLASNAME(cblas_chemm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_cherk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const float alpha, const void *A, const BLASINT lda, + const float beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_cher2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const float beta, + void *C, const BLASINT ldc); + +void BLASNAME(cblas_zhemm)(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, + const enum CBLAS_UPLO Uplo, const BLASINT M, const BLASINT N, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const void *beta, + void *C, const BLASINT ldc); +void BLASNAME(cblas_zherk)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const double alpha, const void *A, const BLASINT lda, + const double beta, void *C, const BLASINT ldc); +void BLASNAME(cblas_zher2k)(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, + const enum CBLAS_TRANSPOSE Trans, const BLASINT N, const BLASINT K, + const void *alpha, const void *A, const BLASINT lda, + const void *B, const BLASINT ldb, const double beta, + void *C, const BLASINT ldc); + +void BLASNAME(cblas_xerbla)(BLASINT p, const char *rout, const char *form, ...); diff --git a/numpy/core/src/common/npy_partition.h.src b/numpy/core/src/common/npy_partition.h.src index a22cf911c..97dc2536b 100644 --- a/numpy/core/src/common/npy_partition.h.src +++ b/numpy/core/src/common/npy_partition.h.src @@ -113,9 +113,6 @@ get_argpartition_func(int type, NPY_SELECTKIND which) npy_intp i; npy_intp ntypes = ARRAY_SIZE(_part_map); - if (which >= NPY_NSELECTS) { - return NULL; - } for (i = 0; i < ntypes; i++) { if (type == _part_map[i].typenum) { return _part_map[i].argpart[which]; diff --git a/numpy/core/src/common/python_xerbla.c b/numpy/core/src/common/python_xerbla.c index bdf0b9058..fe2f718b2 100644 --- a/numpy/core/src/common/python_xerbla.c +++ b/numpy/core/src/common/python_xerbla.c @@ -1,10 +1,6 @@ #include "Python.h" - -/* - * From f2c.h, this should be safe unless fortran is set to use 64 - * bit integers. We don't seem to have any good way to detect that. - */ -typedef int integer; +#include "numpy/npy_common.h" +#include "npy_cblas.h" /* From the original manpage: @@ -23,7 +19,7 @@ typedef int integer; info: Number of the invalid parameter. */ -int xerbla_(char *srname, integer *info) +CBLAS_INT BLAS_FUNC(xerbla)(char *srname, CBLAS_INT *info) { static const char format[] = "On entry to %.*s" \ " parameter number %d had an illegal value"; @@ -41,7 +37,7 @@ int xerbla_(char *srname, integer *info) #ifdef WITH_THREAD save = PyGILState_Ensure(); #endif - PyOS_snprintf(buf, sizeof(buf), format, len, srname, *info); + PyOS_snprintf(buf, sizeof(buf), format, len, srname, (int)*info); PyErr_SetString(PyExc_ValueError, buf); #ifdef WITH_THREAD PyGILState_Release(save); diff --git a/numpy/core/src/common/ufunc_override.c b/numpy/core/src/common/ufunc_override.c index 89f08a9cb..3f699bcdd 100644 --- a/numpy/core/src/common/ufunc_override.c +++ b/numpy/core/src/common/ufunc_override.c @@ -36,6 +36,9 @@ PyUFuncOverride_GetNonDefaultArrayUfunc(PyObject *obj) */ cls_array_ufunc = PyArray_LookupSpecial(obj, "__array_ufunc__"); if (cls_array_ufunc == NULL) { + if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } return NULL; } /* Ignore if the same as ndarray.__array_ufunc__ */ diff --git a/numpy/core/src/multiarray/_multiarray_tests.c.src b/numpy/core/src/multiarray/_multiarray_tests.c.src index 9061c0518..fa2efb428 100644 --- a/numpy/core/src/multiarray/_multiarray_tests.c.src +++ b/numpy/core/src/multiarray/_multiarray_tests.c.src @@ -593,6 +593,25 @@ fail: return NULL; } +/* + * Helper to test fromstring of 0 terminated strings, as the C-API supports + * the -1 length identifier. + */ +static PyObject * +fromstring_null_term_c_api(PyObject *dummy, PyObject *byte_obj) +{ + char *string; + PyArray_Descr *descr; + + string = PyBytes_AsString(byte_obj); + if (string == NULL) { + return NULL; + } + descr = PyArray_DescrNewFromType(NPY_FLOAT64); + return PyArray_FromString(string, -1, descr, -1, " "); +} + + /* check no elison for avoided increfs */ static PyObject * incref_elide(PyObject *dummy, PyObject *args) @@ -656,6 +675,43 @@ npy_updateifcopy_deprecation(PyObject* NPY_UNUSED(self), PyObject* args) Py_RETURN_NONE; } +/* used to test PyArray_As1D usage emits not implemented error */ +static PyObject* +npy_pyarrayas1d_deprecation(PyObject* NPY_UNUSED(self), PyObject* NPY_UNUSED(args)) +{ + PyObject *op = Py_BuildValue("i", 42); + PyObject *result = op; + int dim = 4; + double arg[2] = {1, 2}; + int temp = PyArray_As1D(&result, (char **)&arg, &dim, NPY_DOUBLE); + if (temp < 0) { + Py_DECREF(op); + return NULL; + } + /* op != result */ + Py_DECREF(op); + return result; +} + +/* used to test PyArray_As2D usage emits not implemented error */ +static PyObject* +npy_pyarrayas2d_deprecation(PyObject* NPY_UNUSED(self), PyObject* NPY_UNUSED(args)) +{ + PyObject *op = Py_BuildValue("i", 42); + PyObject *result = op; + int dim1 = 4; + int dim2 = 6; + double arg[2][2] = {{1, 2}, {3, 4}}; + int temp = PyArray_As2D(&result, (char ***)&arg, &dim1, &dim2, NPY_DOUBLE); + if (temp < 0) { + Py_DECREF(op); + return NULL; + } + /* op != result */ + Py_DECREF(op); + return result; +} + /* used to create array with WRITEBACKIFCOPY flag */ static PyObject* npy_create_writebackifcopy(PyObject* NPY_UNUSED(self), PyObject* args) @@ -857,6 +913,7 @@ static PyObject* get_c_wrapping_array(PyObject* NPY_UNUSED(self), PyObject* arg) { int writeable, flags; + PyArray_Descr *descr; npy_intp zero = 0; writeable = PyObject_IsTrue(arg); @@ -866,7 +923,8 @@ get_c_wrapping_array(PyObject* NPY_UNUSED(self), PyObject* arg) flags = writeable ? NPY_ARRAY_WRITEABLE : 0; /* Create an empty array (which points to a random place) */ - return PyArray_NewFromDescr(&PyArray_Type, PyArray_DescrFromType(NPY_INTP), + descr = PyArray_DescrNewFromType(NPY_INTP); + return PyArray_NewFromDescr(&PyArray_Type, descr, 1, &zero, NULL, &zero, flags, NULL); } @@ -928,6 +986,7 @@ test_as_c_array(PyObject *NPY_UNUSED(self), PyObject *args) num_dims = PyArray_NDIM(array_obj); descr = PyArray_DESCR(array_obj); + Py_INCREF(descr); /* PyArray_AsCArray steals a reference to this */ switch (num_dims) { case 1: @@ -970,6 +1029,7 @@ test_as_c_array(PyObject *NPY_UNUSED(self), PyObject *args) PyArray_Free((PyObject *) array_obj, (void *) array3); break; default: + Py_DECREF(descr); PyErr_SetString(PyExc_ValueError, "array.ndim not in [1, 3]"); return NULL; } @@ -1263,7 +1323,9 @@ pylong_from_int128(npy_extint128_t value) } Py_DECREF(val); + Py_DECREF(val_64); val = tmp; + val_64 = NULL; tmp = PyLong_FromUnsignedLongLong(value.lo); if (tmp == NULL) { @@ -1923,6 +1985,9 @@ static PyMethodDef Multiarray_TestsMethods[] = { {"test_inplace_increment", inplace_increment, METH_VARARGS, NULL}, + {"fromstring_null_term_c_api", + fromstring_null_term_c_api, + METH_O, NULL}, {"incref_elide", incref_elide, METH_VARARGS, NULL}, @@ -1935,6 +2000,12 @@ static PyMethodDef Multiarray_TestsMethods[] = { {"npy_updateifcopy_deprecation", npy_updateifcopy_deprecation, METH_O, NULL}, + {"npy_pyarrayas1d_deprecation", + npy_pyarrayas1d_deprecation, + METH_NOARGS, NULL}, + {"npy_pyarrayas2d_deprecation", + npy_pyarrayas2d_deprecation, + METH_NOARGS, NULL}, {"npy_create_writebackifcopy", npy_create_writebackifcopy, METH_O, NULL}, diff --git a/numpy/core/src/multiarray/arrayfunction_override.c b/numpy/core/src/multiarray/arrayfunction_override.c index 62e597764..9ea8efdd9 100644 --- a/numpy/core/src/multiarray/arrayfunction_override.c +++ b/numpy/core/src/multiarray/arrayfunction_override.c @@ -26,6 +26,7 @@ static PyObject * get_array_function(PyObject *obj) { static PyObject *ndarray_array_function = NULL; + PyObject *array_function; if (ndarray_array_function == NULL) { ndarray_array_function = get_ndarray_array_function(); @@ -37,7 +38,12 @@ get_array_function(PyObject *obj) return ndarray_array_function; } - return PyArray_LookupSpecial(obj, "__array_function__"); + array_function = PyArray_LookupSpecial(obj, "__array_function__"); + if (array_function == NULL && PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } + + return array_function; } diff --git a/numpy/core/src/multiarray/arrayobject.c b/numpy/core/src/multiarray/arrayobject.c index ebcb9b0b0..a5cebfbd8 100644 --- a/numpy/core/src/multiarray/arrayobject.c +++ b/numpy/core/src/multiarray/arrayobject.c @@ -462,7 +462,7 @@ WARN_IN_DEALLOC(PyObject* warning, const char * msg) { PyErr_WriteUnraisable(Py_None); } } -}; +} /* array object functions */ @@ -483,10 +483,11 @@ array_dealloc(PyArrayObject *self) char const * msg = "WRITEBACKIFCOPY detected in array_dealloc. " " Required call to PyArray_ResolveWritebackIfCopy or " "PyArray_DiscardWritebackIfCopy is missing."; - Py_INCREF(self); /* hold on to self in next call since if - * refcount == 0 it will recurse back into - *array_dealloc - */ + /* + * prevent reaching 0 twice and thus recursing into dealloc. + * Increasing sys.gettotalrefcount, but path should not be taken. + */ + Py_INCREF(self); WARN_IN_DEALLOC(PyExc_RuntimeWarning, msg); retval = PyArray_ResolveWritebackIfCopy(self); if (retval < 0) @@ -500,10 +501,11 @@ array_dealloc(PyArrayObject *self) char const * msg = "UPDATEIFCOPY detected in array_dealloc. " " Required call to PyArray_ResolveWritebackIfCopy or " "PyArray_DiscardWritebackIfCopy is missing"; - Py_INCREF(self); /* hold on to self in next call since if - * refcount == 0 it will recurse back into - *array_dealloc - */ + /* + * prevent reaching 0 twice and thus recursing into dealloc. + * Increasing sys.gettotalrefcount, but path should not be taken. + */ + Py_INCREF(self); /* 2017-Nov-10 1.14 */ WARN_IN_DEALLOC(PyExc_DeprecationWarning, msg); retval = PyArray_ResolveWritebackIfCopy(self); @@ -523,12 +525,7 @@ array_dealloc(PyArrayObject *self) if ((fa->flags & NPY_ARRAY_OWNDATA) && fa->data) { /* Free internal references if an Object array */ if (PyDataType_FLAGCHK(fa->descr, NPY_ITEM_REFCOUNT)) { - Py_INCREF(self); /*hold on to self */ PyArray_XDECREF(self); - /* - * Don't need to DECREF -- because we are deleting - * self already... - */ } npy_free_cache(fa->data, PyArray_NBYTES(self)); } @@ -560,7 +557,7 @@ PyArray_DebugPrint(PyArrayObject *obj) printf(" ndim : %d\n", fobj->nd); printf(" shape :"); for (i = 0; i < fobj->nd; ++i) { - printf(" %d", (int)fobj->dimensions[i]); + printf(" %" NPY_INTP_FMT, fobj->dimensions[i]); } printf("\n"); @@ -570,7 +567,7 @@ PyArray_DebugPrint(PyArrayObject *obj) printf(" data : %p\n", fobj->data); printf(" strides:"); for (i = 0; i < fobj->nd; ++i) { - printf(" %d", (int)fobj->strides[i]); + printf(" %" NPY_INTP_FMT, fobj->strides[i]); } printf("\n"); @@ -610,7 +607,7 @@ PyArray_DebugPrint(PyArrayObject *obj) * TO BE REMOVED - NOT USED INTERNALLY. */ NPY_NO_EXPORT void -PyArray_SetDatetimeParseFunction(PyObject *op) +PyArray_SetDatetimeParseFunction(PyObject *NPY_UNUSED(op)) { } @@ -633,7 +630,7 @@ PyArray_CompareUCS4(npy_ucs4 *s1, npy_ucs4 *s2, size_t len) /*NUMPY_API */ NPY_NO_EXPORT int -PyArray_CompareString(char *s1, char *s2, size_t len) +PyArray_CompareString(const char *s1, const char *s2, size_t len) { const unsigned char *c1 = (unsigned char *)s1; const unsigned char *c2 = (unsigned char *)s2; @@ -1203,15 +1200,28 @@ _void_compare(PyArrayObject *self, PyArrayObject *other, int cmp_op) } } if (res == NULL && !PyErr_Occurred()) { - PyErr_SetString(PyExc_ValueError, "No fields found."); + /* these dtypes had no fields. Use a MultiIter to broadcast them + * to an output array, and fill with True (for EQ)*/ + PyArrayMultiIterObject *mit = (PyArrayMultiIterObject *) + PyArray_MultiIterNew(2, self, other); + if (mit == NULL) { + return NULL; + } + + res = PyArray_NewFromDescr(&PyArray_Type, + PyArray_DescrFromType(NPY_BOOL), + mit->nd, mit->dimensions, + NULL, NULL, 0, NULL); + Py_DECREF(mit); + if (res) { + PyArray_FILLWBYTE((PyArrayObject *)res, + cmp_op == Py_EQ ? 1 : 0); + } } return res; } else { - /* - * compare as a string. Assumes self and - * other have same descr->type - */ + /* compare as a string. Assumes self and other have same descr->type */ return _strings_richcompare(self, other, cmp_op, 0); } } diff --git a/numpy/core/src/multiarray/arraytypes.c.src b/numpy/core/src/multiarray/arraytypes.c.src index ef51112d6..9e108e3e1 100644 --- a/numpy/core/src/multiarray/arraytypes.c.src +++ b/numpy/core/src/multiarray/arraytypes.c.src @@ -221,9 +221,7 @@ static int if (PySequence_NoString_Check(op)) { PyErr_SetString(PyExc_ValueError, "setting an array element with a sequence."); - Py_DECREF(type); - Py_XDECREF(value); - Py_XDECREF(traceback); + npy_PyErr_ChainExceptionsCause(type, value, traceback); } else { PyErr_Restore(type, value, traceback); @@ -1083,6 +1081,7 @@ TIMEDELTA_setitem(PyObject *op, void *ov, void *vap) * npy_long, npy_ulong, npy_longlong, npy_ulonglong, * npy_float, npy_double, npy_longdouble, * npy_datetime, npy_timedelta# + * #supports_nat = 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1# */ /**begin repeat1 @@ -1094,6 +1093,7 @@ TIMEDELTA_setitem(PyObject *op, void *ov, void *vap) * npy_long, npy_ulong, npy_longlong, npy_ulonglong, * npy_float, npy_double, npy_longdouble, * npy_datetime, npy_timedelta# + * #floatingpoint = 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0# */ static void @FROMTYPE@_to_@TOTYPE@(void *input, void *output, npy_intp n, @@ -1103,7 +1103,15 @@ static void @totype@ *op = output; while (n--) { - *op++ = (@totype@)*ip++; + @fromtype@ f = *ip++; + @totype@ t = (@totype@)f; +#if @supports_nat@ && @floatingpoint@ + /* Avoid undefined behaviour for NaN -> NaT */ + if (npy_isnan(f)) { + t = (@totype@)NPY_DATETIME_NAT; + } +#endif + *op++ = t; } } /**end repeat1**/ @@ -1121,7 +1129,15 @@ static void @totype@ *op = output; while (n--) { - *op++ = (@totype@)*ip; + @fromtype@ f = *ip; + @totype@ t = (@totype@)f; +#if @supports_nat@ + /* Avoid undefined behaviour for NaN -> NaT */ + if (npy_isnan(f)) { + t = (@totype@)NPY_DATETIME_NAT; + } +#endif + *op++ = t; ip += 2; } } @@ -1759,7 +1775,58 @@ BOOL_scan(FILE *fp, npy_bool *ip, void *NPY_UNUSED(ignore), } /**begin repeat - * #fname = CFLOAT, CDOUBLE, CLONGDOUBLE, + * #fname = CFLOAT, CDOUBLE# + * #type = npy_cfloat, npy_cdouble# + */ +static int +@fname@_scan(FILE *fp, @type@ *ip, void *NPY_UNUSED(ignore), + PyArray_Descr *NPY_UNUSED(ignored)) +{ + double result; + int ret_real, ret_imag; + + ret_real = NumPyOS_ascii_ftolf(fp, &result); + @type@ output; + // Peek next character + char next = getc(fp); + if ((next == '+') || (next == '-')) { + // Imaginary component specified + output.real = result; + // Revert peek and read imaginary component + ungetc(next, fp); + ret_imag = NumPyOS_ascii_ftolf(fp, &result); + // Peak next character + next = getc(fp); + if ((ret_imag == 1) && (next == 'j')) { + // If read is successful and the immediate following char is j + output.imag = result; + } + else { + output.imag = 0; + // Push an invalid char to trigger the not everything is read error + ungetc('a', fp); + } + } + else if (next == 'j') { + // Real component not specified + output.real = 0; + output.imag = result; + } + else { + // Imaginary component not specified + output.real = result; + output.imag = 0.; + // Next character is not + / - / j. Revert peek. + ungetc(next, fp); + } + *(@type@ *)ip = output; + return ret_real; +} +/**end repeat**/ + + +/**begin repeat + * #fname = CLONGDOUBLE, * OBJECT, STRING, UNICODE, VOID, * DATETIME, TIMEDELTA# */ @@ -1851,7 +1918,60 @@ BOOL_fromstr(char *str, void *ip, char **endptr, } /**begin repeat - * #fname = CFLOAT, CDOUBLE, CLONGDOUBLE, + * #fname = CFLOAT, CDOUBLE# + * #type = npy_cfloat, npy_cdouble# + */ +static int +@fname@_fromstr(char *str, void *ip, char **endptr, + PyArray_Descr *NPY_UNUSED(ignore)) +{ + double result; + + result = NumPyOS_ascii_strtod(str, endptr); + @type@ output; + + if (endptr && ((*endptr[0] == '+') || (*endptr[0] == '-'))) { + // Imaginary component specified + output.real = result; + // Reading imaginary component + char **prev = endptr; + str = *endptr; + result = NumPyOS_ascii_strtod(str, endptr); + if (endptr && *endptr[0] == 'j') { + // Read is successful if the immediate following char is j + output.imag = result; + // Skip j + ++*endptr; + } + else { + /* + * Set endptr to previous char to trigger the not everything is + * read error + */ + endptr = prev; + output.imag = 0; + } + } + else if (endptr && *endptr[0] == 'j') { + // Real component not specified + output.real = 0; + output.imag = result; + // Skip j + ++*endptr; + } + else { + // Imaginary component not specified + output.real = result; + output.imag = 0.; + } + *(@type@ *)ip = output; + return 0; +} +/**end repeat**/ + + +/**begin repeat + * #fname = CLONGDOUBLE, * OBJECT, STRING, UNICODE, VOID# */ @@ -3080,6 +3200,7 @@ BOOL_argmax(npy_bool *ip, npy_intp n, npy_intp *max_ind, * #le = _LESS_THAN_OR_EQUAL*10, npy_half_le, _LESS_THAN_OR_EQUAL*8# * #iscomplex = 0*14, 1*3, 0*2# * #incr = ip++*14, ip+=2*3, ip++*2# + * #isdatetime = 0*17, 1*2# */ static int @fname@_argmax(@type@ *ip, npy_intp n, npy_intp *max_ind, @@ -3105,6 +3226,12 @@ static int return 0; } #endif +#if @isdatetime@ + if (mp == NPY_DATETIME_NAT) { + /* NaT encountered, it's maximal */ + return 0; + } +#endif for (i = 1; i < n; i++) { @incr@; @@ -3124,6 +3251,13 @@ static int } } #else +#if @isdatetime@ + if (*ip == NPY_DATETIME_NAT) { + /* NaT encountered, it's maximal */ + *max_ind = i; + break; + } +#endif if (!@le@(*ip, mp)) { /* negated, for correct nan handling */ mp = *ip; *max_ind = i; @@ -3160,16 +3294,19 @@ BOOL_argmin(npy_bool *ip, npy_intp n, npy_intp *min_ind, * #fname = BYTE, UBYTE, SHORT, USHORT, INT, UINT, * LONG, ULONG, LONGLONG, ULONGLONG, * HALF, FLOAT, DOUBLE, LONGDOUBLE, - * CFLOAT, CDOUBLE, CLONGDOUBLE# + * CFLOAT, CDOUBLE, CLONGDOUBLE, + * DATETIME, TIMEDELTA# * #type = npy_byte, npy_ubyte, npy_short, npy_ushort, npy_int, npy_uint, * npy_long, npy_ulong, npy_longlong, npy_ulonglong, * npy_half, npy_float, npy_double, npy_longdouble, - * npy_float, npy_double, npy_longdouble# - * #isfloat = 0*10, 1*7# - * #isnan = nop*10, npy_half_isnan, npy_isnan*6# - * #le = _LESS_THAN_OR_EQUAL*10, npy_half_le, _LESS_THAN_OR_EQUAL*6# - * #iscomplex = 0*14, 1*3# - * #incr = ip++*14, ip+=2*3# + * npy_float, npy_double, npy_longdouble, + * npy_datetime, npy_timedelta# + * #isfloat = 0*10, 1*7, 0*2# + * #isnan = nop*10, npy_half_isnan, npy_isnan*6, nop*2# + * #le = _LESS_THAN_OR_EQUAL*10, npy_half_le, _LESS_THAN_OR_EQUAL*8# + * #iscomplex = 0*14, 1*3, 0*2# + * #incr = ip++*14, ip+=2*3, ip++*2# + * #isdatetime = 0*17, 1*2# */ static int @fname@_argmin(@type@ *ip, npy_intp n, npy_intp *min_ind, @@ -3195,6 +3332,12 @@ static int return 0; } #endif +#if @isdatetime@ + if (mp == NPY_DATETIME_NAT) { + /* NaT encountered, it's minimal */ + return 0; + } +#endif for (i = 1; i < n; i++) { @incr@; @@ -3214,6 +3357,13 @@ static int } } #else +#if @isdatetime@ + if (*ip == NPY_DATETIME_NAT) { + /* NaT encountered, it's minimal */ + *min_ind = i; + break; + } +#endif if (!@le@(mp, *ip)) { /* negated, for correct nan handling */ mp = *ip; *min_ind = i; @@ -3233,43 +3383,6 @@ static int #undef _LESS_THAN_OR_EQUAL -/**begin repeat - * - * #fname = DATETIME, TIMEDELTA# - * #type = npy_datetime, npy_timedelta# - */ -static int -@fname@_argmin(@type@ *ip, npy_intp n, npy_intp *min_ind, - PyArrayObject *NPY_UNUSED(aip)) -{ - /* NPY_DATETIME_NAT is smaller than every other value, we skip - * it for consistency with min(). - */ - npy_intp i; - @type@ mp = NPY_DATETIME_NAT; - - i = 0; - while (i < n && mp == NPY_DATETIME_NAT) { - mp = ip[i]; - i++; - } - if (i == n) { - /* All NaTs: return 0 */ - *min_ind = 0; - return 0; - } - *min_ind = i - 1; - for (; i < n; i++) { - if (mp > ip[i] && ip[i] != NPY_DATETIME_NAT) { - mp = ip[i]; - *min_ind = i; - } - } - return 0; -} - -/**end repeat**/ - static int OBJECT_argmax(PyObject **ip, npy_intp n, npy_intp *max_ind, PyArrayObject *NPY_UNUSED(aip)) @@ -3422,17 +3535,17 @@ NPY_NO_EXPORT void npy_intp n, void *NPY_UNUSED(ignore)) { #if defined(HAVE_CBLAS) - int is1b = blas_stride(is1, sizeof(@type@)); - int is2b = blas_stride(is2, sizeof(@type@)); + CBLAS_INT is1b = blas_stride(is1, sizeof(@type@)); + CBLAS_INT is2b = blas_stride(is2, sizeof(@type@)); if (is1b && is2b) { double sum = 0.; /* double for stability */ while (n > 0) { - int chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; + CBLAS_INT chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; - sum += cblas_@prefix@dot(chunk, + sum += CBLAS_FUNC(cblas_@prefix@dot)(chunk, (@type@ *) ip1, is1b, (@type@ *) ip2, is2b); /* use char strides here */ @@ -3471,17 +3584,17 @@ NPY_NO_EXPORT void char *op, npy_intp n, void *NPY_UNUSED(ignore)) { #if defined(HAVE_CBLAS) - int is1b = blas_stride(is1, sizeof(@ctype@)); - int is2b = blas_stride(is2, sizeof(@ctype@)); + CBLAS_INT is1b = blas_stride(is1, sizeof(@ctype@)); + CBLAS_INT is2b = blas_stride(is2, sizeof(@ctype@)); if (is1b && is2b) { double sum[2] = {0., 0.}; /* double for stability */ while (n > 0) { - int chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; + CBLAS_INT chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; @type@ tmp[2]; - cblas_@prefix@dotu_sub((int)n, ip1, is1b, ip2, is2b, tmp); + CBLAS_FUNC(cblas_@prefix@dotu_sub)((CBLAS_INT)n, ip1, is1b, ip2, is2b, tmp); sum[0] += (double)tmp[0]; sum[1] += (double)tmp[1]; /* use char strides here */ diff --git a/numpy/core/src/multiarray/common.c b/numpy/core/src/multiarray/common.c index 6b8cac9a7..c991f7428 100644 --- a/numpy/core/src/multiarray/common.c +++ b/numpy/core/src/multiarray/common.c @@ -147,7 +147,6 @@ PyArray_DTypeFromObjectHelper(PyObject *obj, int maxdims, if (dtype == NULL) { goto fail; } - Py_INCREF(dtype); goto promote_types; } /* Check if it's a NumPy scalar */ @@ -214,6 +213,10 @@ PyArray_DTypeFromObjectHelper(PyObject *obj, int maxdims, int itemsize; PyObject *temp; + /* dtype is not used in this (string discovery) branch */ + Py_DECREF(dtype); + dtype = NULL; + if (string_type == NPY_STRING) { if ((temp = PyObject_Str(obj)) == NULL) { goto fail; @@ -364,6 +367,10 @@ PyArray_DTypeFromObjectHelper(PyObject *obj, int maxdims, } Py_DECREF(ip); } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } + /* The array struct interface */ ip = PyArray_LookupSpecial_OnInstance(obj, "__array_struct__"); @@ -386,6 +393,9 @@ PyArray_DTypeFromObjectHelper(PyObject *obj, int maxdims, } Py_DECREF(ip); } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } /* The old buffer interface */ #if !defined(NPY_PY3K) @@ -416,6 +426,9 @@ PyArray_DTypeFromObjectHelper(PyObject *obj, int maxdims, goto fail; } } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } /* * If we reached the maximum recursion depth without hitting one diff --git a/numpy/core/src/multiarray/common.h b/numpy/core/src/multiarray/common.h index 487d530a1..7eee9ddc5 100644 --- a/numpy/core/src/multiarray/common.h +++ b/numpy/core/src/multiarray/common.h @@ -303,7 +303,11 @@ blas_stride(npy_intp stride, unsigned itemsize) */ if (stride > 0 && npy_is_aligned((void *)stride, itemsize)) { stride /= itemsize; +#ifndef HAVE_BLAS_ILP64 if (stride <= INT_MAX) { +#else + if (stride <= NPY_MAX_INT64) { +#endif return stride; } } @@ -314,7 +318,11 @@ blas_stride(npy_intp stride, unsigned itemsize) * Define a chunksize for CBLAS. CBLAS counts in integers. */ #if NPY_MAX_INTP > INT_MAX -# define NPY_CBLAS_CHUNK (INT_MAX / 2 + 1) +# ifndef HAVE_BLAS_ILP64 +# define NPY_CBLAS_CHUNK (INT_MAX / 2 + 1) +# else +# define NPY_CBLAS_CHUNK (NPY_MAX_INT64 / 2 + 1) +# endif #else # define NPY_CBLAS_CHUNK NPY_MAX_INTP #endif diff --git a/numpy/core/src/multiarray/compiled_base.c b/numpy/core/src/multiarray/compiled_base.c index c38067681..055d3e60f 100644 --- a/numpy/core/src/multiarray/compiled_base.c +++ b/numpy/core/src/multiarray/compiled_base.c @@ -942,6 +942,20 @@ ravel_multi_index_loop(int ravel_ndim, npy_intp *ravel_dims, char invalid; npy_intp j, m; + /* + * Check for 0-dimensional axes unless there is nothing to do. + * An empty array/shape cannot be indexed at all. + */ + if (count != 0) { + for (i = 0; i < ravel_ndim; ++i) { + if (ravel_dims[i] == 0) { + PyErr_SetString(PyExc_ValueError, + "cannot unravel if shape has zero entries (is empty)."); + return NPY_FAIL; + } + } + } + NPY_BEGIN_ALLOW_THREADS; invalid = 0; while (count--) { diff --git a/numpy/core/src/multiarray/conversion_utils.c b/numpy/core/src/multiarray/conversion_utils.c index 52cb58726..ca126b4b1 100644 --- a/numpy/core/src/multiarray/conversion_utils.c +++ b/numpy/core/src/multiarray/conversion_utils.c @@ -406,7 +406,6 @@ PyArray_SortkindConverter(PyObject *obj, NPY_SORTKIND *sortkind) } *sortkind = NPY_QUICKSORT; - str = PyBytes_AsString(obj); if (!str) { @@ -551,10 +550,9 @@ PyArray_OrderConverter(PyObject *object, NPY_ORDER *val) int ret; tmp = PyUnicode_AsASCIIString(object); if (tmp == NULL) { - PyErr_SetString(PyExc_ValueError, "Invalid unicode string passed in " - "for the array ordering. " - "Please pass in 'C', 'F', 'A' " - "or 'K' instead"); + PyErr_SetString(PyExc_ValueError, + "Invalid unicode string passed in for the array ordering. " + "Please pass in 'C', 'F', 'A' or 'K' instead"); return NPY_FAIL; } ret = PyArray_OrderConverter(tmp, val); @@ -562,38 +560,18 @@ PyArray_OrderConverter(PyObject *object, NPY_ORDER *val) return ret; } else if (!PyBytes_Check(object) || PyBytes_GET_SIZE(object) < 1) { - /* 2015-12-14, 1.11 */ - int ret = DEPRECATE("Non-string object detected for " - "the array ordering. Please pass " - "in 'C', 'F', 'A', or 'K' instead"); - - if (ret < 0) { - return -1; - } - - if (PyObject_IsTrue(object)) { - *val = NPY_FORTRANORDER; - } - else { - *val = NPY_CORDER; - } - if (PyErr_Occurred()) { - return NPY_FAIL; - } - return NPY_SUCCEED; + PyErr_SetString(PyExc_ValueError, + "Non-string object detected for the array ordering. " + "Please pass in 'C', 'F', 'A', or 'K' instead"); + return NPY_FAIL; } else { str = PyBytes_AS_STRING(object); if (strlen(str) != 1) { - /* 2015-12-14, 1.11 */ - int ret = DEPRECATE("Non length-one string passed " - "in for the array ordering. " - "Please pass in 'C', 'F', 'A', " - "or 'K' instead"); - - if (ret < 0) { - return -1; - } + PyErr_SetString(PyExc_ValueError, + "Non-string object detected for the array ordering. " + "Please pass in 'C', 'F', 'A', or 'K' instead"); + return NPY_FAIL; } if (str[0] == 'C' || str[0] == 'c') { @@ -689,8 +667,8 @@ PyArray_ConvertClipmodeSequence(PyObject *object, NPY_CLIPMODE *modes, int n) if (object && (PyTuple_Check(object) || PyList_Check(object))) { if (PySequence_Size(object) != n) { PyErr_Format(PyExc_ValueError, - "list of clipmodes has wrong length (%d instead of %d)", - (int)PySequence_Size(object), n); + "list of clipmodes has wrong length (%zd instead of %d)", + PySequence_Size(object), n); return NPY_FAIL; } diff --git a/numpy/core/src/multiarray/convert_datatype.c b/numpy/core/src/multiarray/convert_datatype.c index 025c66013..4326448dc 100644 --- a/numpy/core/src/multiarray/convert_datatype.c +++ b/numpy/core/src/multiarray/convert_datatype.c @@ -877,7 +877,13 @@ PyArray_CanCastTypeTo(PyArray_Descr *from, PyArray_Descr *to, from_order = dtype_kind_to_ordering(from->kind); to_order = dtype_kind_to_ordering(to->kind); - return from_order != -1 && from_order <= to_order; + if (to->kind == 'm') { + /* both types being timedelta is already handled before. */ + int integer_order = dtype_kind_to_ordering('i'); + return (from_order != -1) && (from_order <= integer_order); + } + + return (from_order != -1) && (from_order <= to_order); } else { return 0; diff --git a/numpy/core/src/multiarray/ctors.c b/numpy/core/src/multiarray/ctors.c index 0897feaf4..7276add75 100644 --- a/numpy/core/src/multiarray/ctors.c +++ b/numpy/core/src/multiarray/ctors.c @@ -40,9 +40,31 @@ * regards to the handling of text representations. */ +/* + * Scanning function for next element parsing and seperator skipping. + * These functions return: + * - 0 to indicate more data to read + * - -1 when reading stopped at the end of the string/file + * - -2 when reading stopped before the end was reached. + * + * The dtype specific parsing functions may set the python error state + * (they have to get the GIL first) additionally. + */ typedef int (*next_element)(void **, void *, PyArray_Descr *, void *); typedef int (*skip_separator)(void **, const char *, void *); + +static npy_bool +string_is_fully_read(char const* start, char const* end) { + if (end == NULL) { + return *start == '\0'; /* null terminated */ + } + else { + return start >= end; /* fixed length */ + } +} + + static int fromstr_next_element(char **s, void *dptr, PyArray_Descr *dtype, const char *end) @@ -50,19 +72,23 @@ fromstr_next_element(char **s, void *dptr, PyArray_Descr *dtype, char *e = *s; int r = dtype->f->fromstr(*s, dptr, &e, dtype); /* - * fromstr always returns 0 for basic dtypes - * s points to the end of the parsed string - * if an error occurs s is not changed + * fromstr always returns 0 for basic dtypes; s points to the end of the + * parsed string. If s is not changed an error occurred or the end was + * reached. */ - if (*s == e) { - /* Nothing read */ - return -1; + if (*s == e || r < 0) { + /* Nothing read, could be end of string or an error (or both) */ + if (string_is_fully_read(*s, end)) { + return -1; + } + return -2; } *s = e; if (end != NULL && *s > end) { + /* Stop the iteration if we read far enough */ return -1; } - return r; + return 0; } static int @@ -75,9 +101,13 @@ fromfile_next_element(FILE **fp, void *dptr, PyArray_Descr *dtype, if (r == 1) { return 0; } - else { + else if (r == EOF) { return -1; } + else { + /* unable to read more, but EOF not reached indicating an error. */ + return -2; + } } /* @@ -143,9 +173,10 @@ fromstr_skip_separator(char **s, const char *sep, const char *end) { char *string = *s; int result = 0; + while (1) { char c = *string; - if (c == '\0' || (end != NULL && string >= end)) { + if (string_is_fully_read(string, end)) { result = -1; break; } @@ -488,8 +519,8 @@ setArrayFromSequence(PyArrayObject *a, PyObject *s, */ if (slen != PyArray_DIMS(a)[dim] && slen != 1) { PyErr_Format(PyExc_ValueError, - "cannot copy sequence with size %d to array axis " - "with dimension %d", (int)slen, (int)PyArray_DIMS(a)[dim]); + "cannot copy sequence with size %zd to array axis " + "with dimension %" NPY_INTP_FMT, slen, PyArray_DIMS(a)[dim]); goto fail; } @@ -796,6 +827,10 @@ discover_dimensions(PyObject *obj, int *maxndim, npy_intp *d, int check_it, return 0; } } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } + /* obj has the __array_interface__ interface */ e = PyArray_LookupSpecial_OnInstance(obj, "__array_interface__"); @@ -825,6 +860,9 @@ discover_dimensions(PyObject *obj, int *maxndim, npy_intp *d, int check_it, return 0; } } + else if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } seq = PySequence_Fast(obj, "Could not convert object to sequence"); if (seq == NULL) { @@ -911,6 +949,39 @@ discover_dimensions(PyObject *obj, int *maxndim, npy_intp *d, int check_it, return 0; } +static PyObject * +raise_memory_error(int nd, npy_intp *dims, PyArray_Descr *descr) +{ + static PyObject *exc_type = NULL; + + npy_cache_import( + "numpy.core._exceptions", "_ArrayMemoryError", + &exc_type); + if (exc_type == NULL) { + goto fail; + } + + PyObject *shape = PyArray_IntTupleFromIntp(nd, dims); + if (shape == NULL) { + goto fail; + } + + /* produce an error object */ + PyObject *exc_value = PyTuple_Pack(2, shape, (PyObject *)descr); + Py_DECREF(shape); + if (exc_value == NULL){ + goto fail; + } + PyErr_SetObject(exc_type, exc_value); + Py_DECREF(exc_value); + return NULL; + +fail: + /* we couldn't raise the formatted exception for some reason */ + PyErr_WriteUnraisable(NULL); + return PyErr_NoMemory(); +} + /* * Generic new array creation routine. * Internal variant with calloc argument for PyArray_Zeros. @@ -1088,30 +1159,7 @@ PyArray_NewFromDescr_int( data = npy_alloc_cache(nbytes); } if (data == NULL) { - static PyObject *exc_type = NULL; - - npy_cache_import( - "numpy.core._exceptions", "_ArrayMemoryError", - &exc_type); - if (exc_type == NULL) { - return NULL; - } - - PyObject *shape = PyArray_IntTupleFromIntp(fa->nd,fa->dimensions); - if (shape == NULL) { - return NULL; - } - - /* produce an error object */ - PyObject *exc_value = PyTuple_Pack(2, shape, descr); - Py_DECREF(shape); - if (exc_value == NULL){ - return NULL; - } - PyErr_SetObject(exc_type, exc_value); - Py_DECREF(exc_value); - return NULL; - + return raise_memory_error(fa->nd, fa->dimensions, descr); } fa->flags |= NPY_ARRAY_OWNDATA; @@ -1873,6 +1921,7 @@ PyArray_FromAny(PyObject *op, PyArray_Descr *newtype, int min_depth, if (arr == NULL) { if ((flags & NPY_ARRAY_WRITEBACKIFCOPY) || (flags & NPY_ARRAY_UPDATEIFCOPY)) { + Py_DECREF(dtype); Py_XDECREF(newtype); PyErr_SetString(PyExc_TypeError, "WRITEBACKIFCOPY used for non-array input."); @@ -2249,7 +2298,11 @@ PyArray_FromStructInterface(PyObject *input) attr = PyArray_LookupSpecial_OnInstance(input, "__array_struct__"); if (attr == NULL) { - return Py_NotImplemented; + if (PyErr_Occurred()) { + return NULL; + } else { + return Py_NotImplemented; + } } if (!NpyCapsule_Check(attr)) { goto fail; @@ -2361,6 +2414,9 @@ PyArray_FromInterface(PyObject *origin) iface = PyArray_LookupSpecial_OnInstance(origin, "__array_interface__"); if (iface == NULL) { + if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } return Py_NotImplemented; } if (!PyDict_Check(iface)) { @@ -2614,6 +2670,9 @@ PyArray_FromArrayAttr(PyObject *op, PyArray_Descr *typecode, PyObject *context) array_meth = PyArray_LookupSpecial_OnInstance(op, "__array__"); if (array_meth == NULL) { + if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } return Py_NotImplemented; } if (context == NULL) { @@ -2682,61 +2741,30 @@ PyArray_DescrFromObject(PyObject *op, PyArray_Descr *mintype) /* They all zero-out the memory as previously done */ /* steals reference to descr -- and enforces native byteorder on it.*/ + /*NUMPY_API - Like FromDimsAndData but uses the Descr structure instead of typecode - as input. + Deprecated, use PyArray_NewFromDescr instead. */ NPY_NO_EXPORT PyObject * -PyArray_FromDimsAndDataAndDescr(int nd, int *d, +PyArray_FromDimsAndDataAndDescr(int NPY_UNUSED(nd), int *NPY_UNUSED(d), PyArray_Descr *descr, - char *data) + char *NPY_UNUSED(data)) { - PyObject *ret; - int i; - npy_intp newd[NPY_MAXDIMS]; - char msg[] = "PyArray_FromDimsAndDataAndDescr: use PyArray_NewFromDescr."; - - if (DEPRECATE(msg) < 0) { - /* 2009-04-30, 1.5 */ - return NULL; - } - if (!PyArray_ISNBO(descr->byteorder)) - descr->byteorder = '='; - for (i = 0; i < nd; i++) { - newd[i] = (npy_intp) d[i]; - } - ret = PyArray_NewFromDescr(&PyArray_Type, descr, - nd, newd, - NULL, data, - (data ? NPY_ARRAY_CARRAY : 0), NULL); - return ret; + PyErr_SetString(PyExc_NotImplementedError, + "PyArray_FromDimsAndDataAndDescr: use PyArray_NewFromDescr."); + Py_DECREF(descr); + return NULL; } /*NUMPY_API - Construct an empty array from dimensions and typenum + Deprecated, use PyArray_SimpleNew instead. */ NPY_NO_EXPORT PyObject * -PyArray_FromDims(int nd, int *d, int type) +PyArray_FromDims(int NPY_UNUSED(nd), int *NPY_UNUSED(d), int NPY_UNUSED(type)) { - PyArrayObject *ret; - char msg[] = "PyArray_FromDims: use PyArray_SimpleNew."; - - if (DEPRECATE(msg) < 0) { - /* 2009-04-30, 1.5 */ - return NULL; - } - ret = (PyArrayObject *)PyArray_FromDimsAndDataAndDescr(nd, d, - PyArray_DescrFromType(type), - NULL); - /* - * Old FromDims set memory to zero --- some algorithms - * relied on that. Better keep it the same. If - * Object type, then it's already been set to zero, though. - */ - if (ret && (PyArray_DESCR(ret)->type_num != NPY_OBJECT)) { - memset(PyArray_DATA(ret), 0, PyArray_NBYTES(ret)); - } - return (PyObject *)ret; + PyErr_SetString(PyExc_NotImplementedError, + "PyArray_FromDims: use PyArray_SimpleNew."); + return NULL; } /* end old calls */ @@ -2823,8 +2851,8 @@ PyArray_CopyAsFlat(PyArrayObject *dst, PyArrayObject *src, NPY_ORDER order) src_size = PyArray_SIZE(src); if (dst_size != src_size) { PyErr_Format(PyExc_ValueError, - "cannot copy from array of size %d into an array " - "of size %d", (int)src_size, (int)dst_size); + "cannot copy from array of size %" NPY_INTP_FMT " into an array " + "of size %" NPY_INTP_FMT, src_size, dst_size); return -1; } @@ -3503,11 +3531,13 @@ PyArray_ArangeObj(PyObject *start, PyObject *stop, PyObject *step, PyArray_Descr return NULL; } +/* This array creation function steals the reference to dtype. */ static PyArrayObject * array_fromfile_binary(FILE *fp, PyArray_Descr *dtype, npy_intp num, size_t *nread) { PyArrayObject *r; npy_off_t start, numbytes; + int elsize; if (num < 0) { int fail = 0; @@ -3534,27 +3564,29 @@ array_fromfile_binary(FILE *fp, PyArray_Descr *dtype, npy_intp num, size_t *nrea } num = numbytes / dtype->elsize; } + /* - * When dtype->subarray is true, PyArray_NewFromDescr will decref dtype - * even on success, so make sure it stays around until exit. + * Array creation may move sub-array dimensions from the dtype to array + * dimensions, so we need to use the original element size when reading. */ - Py_INCREF(dtype); + elsize = dtype->elsize; + r = (PyArrayObject *)PyArray_NewFromDescr(&PyArray_Type, dtype, 1, &num, NULL, NULL, 0, NULL); if (r == NULL) { - Py_DECREF(dtype); return NULL; } + NPY_BEGIN_ALLOW_THREADS; - *nread = fread(PyArray_DATA(r), dtype->elsize, num, fp); + *nread = fread(PyArray_DATA(r), elsize, num, fp); NPY_END_ALLOW_THREADS; - Py_DECREF(dtype); return r; } /* * Create an array by reading from the given stream, using the passed * next_element and skip_separator functions. + * As typical for array creation functions, it steals the reference to dtype. */ #define FROM_BUFFER_SIZE 4096 static PyArrayObject * @@ -3566,6 +3598,7 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, npy_intp i; char *dptr, *clean_sep, *tmp; int err = 0; + int stop_reading_flag; /* -1 indicates end reached; -2 a parsing error */ npy_intp thisbuf = 0; npy_intp size; npy_intp bytes, totalbytes; @@ -3573,10 +3606,11 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, size = (num >= 0) ? num : FROM_BUFFER_SIZE; /* - * When dtype->subarray is true, PyArray_NewFromDescr will decref dtype - * even on success, so make sure it stays around until exit. + * Array creation may move sub-array dimensions from the dtype to array + * dimensions, so we need to use the original dtype when reading. */ Py_INCREF(dtype); + r = (PyArrayObject *) PyArray_NewFromDescr(&PyArray_Type, dtype, 1, &size, NULL, NULL, 0, NULL); @@ -3584,6 +3618,7 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, Py_DECREF(dtype); return NULL; } + clean_sep = swab_separator(sep); if (clean_sep == NULL) { err = 1; @@ -3593,9 +3628,9 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, NPY_BEGIN_ALLOW_THREADS; totalbytes = bytes = size * dtype->elsize; dptr = PyArray_DATA(r); - for (i= 0; num < 0 || i < num; i++) { - if (next(&stream, dptr, dtype, stream_data) < 0) { - /* EOF */ + for (i = 0; num < 0 || i < num; i++) { + stop_reading_flag = next(&stream, dptr, dtype, stream_data); + if (stop_reading_flag < 0) { break; } *nread += 1; @@ -3612,7 +3647,12 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, dptr = tmp + (totalbytes - bytes); thisbuf = 0; } - if (skip_sep(&stream, clean_sep, stream_data) < 0) { + stop_reading_flag = skip_sep(&stream, clean_sep, stream_data); + if (stop_reading_flag < 0) { + if (num == i + 1) { + /* if we read as much as requested sep is optional */ + stop_reading_flag = -1; + } break; } } @@ -3631,8 +3671,24 @@ array_from_text(PyArray_Descr *dtype, npy_intp num, char *sep, size_t *nread, } } NPY_END_ALLOW_THREADS; + free(clean_sep); + if (stop_reading_flag == -2) { + if (PyErr_Occurred()) { + /* If an error is already set (unlikely), do not create new one */ + Py_DECREF(r); + Py_DECREF(dtype); + return NULL; + } + /* 2019-09-12, NumPy 1.18 */ + if (DEPRECATE( + "string or file could not be read to its end due to unmatched " + "data; this will raise a ValueError in the future.") < 0) { + goto fail; + } + } + fail: Py_DECREF(dtype); if (err == 1) { @@ -3651,9 +3707,8 @@ fail: * Given a ``FILE *`` pointer ``fp``, and a ``PyArray_Descr``, return an * array corresponding to the data encoded in that file. * - * If the dtype is NULL, the default array type is used (double). - * If non-null, the reference is stolen and if dtype->subarray is true dtype - * will be decrefed even on success. + * The reference to `dtype` is stolen (it is possible that the passed in + * dtype is not held on to). * * The number of elements to read is given as ``num``; if it is < 0, then * then as many as possible are read. @@ -3701,7 +3756,6 @@ PyArray_FromFile(FILE *fp, PyArray_Descr *dtype, npy_intp num, char *sep) (skip_separator) fromfile_skip_separator, NULL); } if (ret == NULL) { - Py_DECREF(dtype); return NULL; } if (((npy_intp) nread) < num) { @@ -3791,7 +3845,13 @@ PyArray_FromBuffer(PyObject *buf, PyArray_Descr *type, s = (npy_intp)ts - offset; n = (npy_intp)count; itemsize = type->elsize; - if (n < 0 ) { + if (n < 0) { + if (itemsize == 0) { + PyErr_SetString(PyExc_ValueError, + "cannot determine count if itemsize is 0"); + Py_DECREF(type); + return NULL; + } if (s % itemsize != 0) { PyErr_SetString(PyExc_ValueError, "buffer size must be a multiple"\ @@ -3896,6 +3956,11 @@ PyArray_FromString(char *data, npy_intp slen, PyArray_Descr *dtype, return NULL; } } + /* + * NewFromDescr may replace dtype to absorb subarray shape + * into the array, so get size beforehand. + */ + npy_intp size_to_copy = num*dtype->elsize; ret = (PyArrayObject *) PyArray_NewFromDescr(&PyArray_Type, dtype, 1, &num, NULL, NULL, @@ -3903,14 +3968,14 @@ PyArray_FromString(char *data, npy_intp slen, PyArray_Descr *dtype, if (ret == NULL) { return NULL; } - memcpy(PyArray_DATA(ret), data, num*dtype->elsize); + memcpy(PyArray_DATA(ret), data, size_to_copy); } else { /* read from character-based string */ size_t nread = 0; char *end; - if (dtype->f->scanfunc == NULL) { + if (dtype->f->fromstr == NULL) { PyErr_SetString(PyExc_ValueError, "don't know how to read " \ "character strings with that " \ @@ -3984,7 +4049,7 @@ PyArray_FromIter(PyObject *obj, PyArray_Descr *dtype, npy_intp count) } for (i = 0; (i < count || count == -1) && (value = PyIter_Next(iter)); i++) { - if (i >= elcount) { + if (i >= elcount && elsize != 0) { npy_intp nbytes; /* Grow PyArray_DATA(ret): diff --git a/numpy/core/src/multiarray/datetime.c b/numpy/core/src/multiarray/datetime.c index 4268b8893..72a3df89c 100644 --- a/numpy/core/src/multiarray/datetime.c +++ b/numpy/core/src/multiarray/datetime.c @@ -27,6 +27,40 @@ #include "datetime_strings.h" /* + * Computes the python `ret, d = divmod(d, unit)`. + * + * Note that GCC is smart enough at -O2 to eliminate the `if(*d < 0)` branch + * for subsequent calls to this command - it is able to deduce that `*d >= 0`. + */ +static inline +npy_int64 extract_unit_64(npy_int64 *d, npy_int64 unit) { + assert(unit > 0); + npy_int64 div = *d / unit; + npy_int64 mod = *d % unit; + if (mod < 0) { + mod += unit; + div -= 1; + } + assert(mod >= 0); + *d = mod; + return div; +} + +static inline +npy_int32 extract_unit_32(npy_int32 *d, npy_int32 unit) { + assert(unit > 0); + npy_int32 div = *d / unit; + npy_int32 mod = *d % unit; + if (mod < 0) { + mod += unit; + div -= 1; + } + assert(mod >= 0); + *d = mod; + return div; +} + +/* * Imports the PyDateTime functions so we can create these objects. * This is called during module initialization */ @@ -160,17 +194,7 @@ days_to_yearsdays(npy_int64 *days_) npy_int64 year; /* Break down the 400 year cycle to get the year and day within the year */ - if (days >= 0) { - year = 400 * (days / days_per_400years); - days = days % days_per_400years; - } - else { - year = 400 * ((days - (days_per_400years - 1)) / days_per_400years); - days = days % days_per_400years; - if (days < 0) { - days += days_per_400years; - } - } + year = 400 * extract_unit_64(&days, days_per_400years); /* Work out the year/day within the 400 year cycle */ if (days >= 366) { @@ -386,7 +410,8 @@ convert_datetimestruct_to_datetime(PyArray_DatetimeMetaData *meta, * TO BE REMOVED - NOT USED INTERNALLY. */ NPY_NO_EXPORT npy_datetime -PyArray_DatetimeStructToDatetime(NPY_DATETIMEUNIT fr, npy_datetimestruct *d) +PyArray_DatetimeStructToDatetime( + NPY_DATETIMEUNIT NPY_UNUSED(fr), npy_datetimestruct *NPY_UNUSED(d)) { PyErr_SetString(PyExc_RuntimeError, "The NumPy PyArray_DatetimeStructToDatetime function has " @@ -400,7 +425,8 @@ PyArray_DatetimeStructToDatetime(NPY_DATETIMEUNIT fr, npy_datetimestruct *d) * TO BE REMOVED - NOT USED INTERNALLY. */ NPY_NO_EXPORT npy_datetime -PyArray_TimedeltaStructToTimedelta(NPY_DATETIMEUNIT fr, npy_timedeltastruct *d) +PyArray_TimedeltaStructToTimedelta( + NPY_DATETIMEUNIT NPY_UNUSED(fr), npy_timedeltastruct *NPY_UNUSED(d)) { PyErr_SetString(PyExc_RuntimeError, "The NumPy PyArray_TimedeltaStructToTimedelta function has " @@ -409,26 +435,6 @@ PyArray_TimedeltaStructToTimedelta(NPY_DATETIMEUNIT fr, npy_timedeltastruct *d) } /* - * Computes the python `ret, d = divmod(d, unit)`. - * - * Note that GCC is smart enough at -O2 to eliminate the `if(*d < 0)` branch - * for subsequent calls to this command - it is able to deduce that `*d >= 0`. - */ -static inline -npy_int64 extract_unit(npy_datetime *d, npy_datetime unit) { - assert(unit > 0); - npy_int64 div = *d / unit; - npy_int64 mod = *d % unit; - if (mod < 0) { - mod += unit; - div -= 1; - } - assert(mod >= 0); - *d = mod; - return div; -} - -/* * Converts a datetime based on the given metadata into a datetimestruct */ NPY_NO_EXPORT int @@ -436,7 +442,7 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, npy_datetime dt, npy_datetimestruct *out) { - npy_int64 perday; + npy_int64 days; /* Initialize the output to all zeros */ memset(out, 0, sizeof(npy_datetimestruct)); @@ -471,7 +477,7 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, break; case NPY_FR_M: - out->year = 1970 + extract_unit(&dt, 12); + out->year = 1970 + extract_unit_64(&dt, 12); out->month = dt + 1; break; @@ -485,73 +491,67 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, break; case NPY_FR_h: - perday = 24LL; - - set_datetimestruct_days(extract_unit(&dt, perday), out); + days = extract_unit_64(&dt, 24LL); + set_datetimestruct_days(days, out); out->hour = (int)dt; break; case NPY_FR_m: - perday = 24LL * 60; - - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 60); - out->min = (int)dt; + days = extract_unit_64(&dt, 60LL*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 60LL); + out->min = (int)dt; break; case NPY_FR_s: - perday = 24LL * 60 * 60; - - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 60*60); - out->min = (int)extract_unit(&dt, 60); + days = extract_unit_64(&dt, 60LL*60*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 60LL*60); + out->min = (int)extract_unit_64(&dt, 60LL); out->sec = (int)dt; break; case NPY_FR_ms: - perday = 24LL * 60 * 60 * 1000; - - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 1000LL*60*60); - out->min = (int)extract_unit(&dt, 1000LL*60); - out->sec = (int)extract_unit(&dt, 1000LL); + days = extract_unit_64(&dt, 1000LL*60*60*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 1000LL*60*60); + out->min = (int)extract_unit_64(&dt, 1000LL*60); + out->sec = (int)extract_unit_64(&dt, 1000LL); out->us = (int)(dt * 1000); break; case NPY_FR_us: - perday = 24LL * 60LL * 60LL * 1000LL * 1000LL; - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 1000LL*1000*60*60); - out->min = (int)extract_unit(&dt, 1000LL*1000*60); - out->sec = (int)extract_unit(&dt, 1000LL*1000); + days = extract_unit_64(&dt, 1000LL*1000*60*60*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 1000LL*1000*60*60); + out->min = (int)extract_unit_64(&dt, 1000LL*1000*60); + out->sec = (int)extract_unit_64(&dt, 1000LL*1000); out->us = (int)dt; break; case NPY_FR_ns: - perday = 24LL * 60LL * 60LL * 1000LL * 1000LL * 1000LL; - - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 1000LL*1000*1000*60*60); - out->min = (int)extract_unit(&dt, 1000LL*1000*1000*60); - out->sec = (int)extract_unit(&dt, 1000LL*1000*1000); - out->us = (int)extract_unit(&dt, 1000LL); + days = extract_unit_64(&dt, 1000LL*1000*1000*60*60*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 1000LL*1000*1000*60*60); + out->min = (int)extract_unit_64(&dt, 1000LL*1000*1000*60); + out->sec = (int)extract_unit_64(&dt, 1000LL*1000*1000); + out->us = (int)extract_unit_64(&dt, 1000LL); out->ps = (int)(dt * 1000); break; case NPY_FR_ps: - perday = 24LL * 60 * 60 * 1000 * 1000 * 1000 * 1000; - - set_datetimestruct_days(extract_unit(&dt, perday), out); - out->hour = (int)extract_unit(&dt, 1000LL*1000*1000*1000*60*60); - out->min = (int)extract_unit(&dt, 1000LL*1000*1000*1000*60); - out->sec = (int)extract_unit(&dt, 1000LL*1000*1000*1000); - out->us = (int)extract_unit(&dt, 1000LL*1000); + days = extract_unit_64(&dt, 1000LL*1000*1000*1000*60*60*24); + set_datetimestruct_days(days, out); + out->hour = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*60*60); + out->min = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*60); + out->sec = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000); + out->us = (int)extract_unit_64(&dt, 1000LL*1000); out->ps = (int)(dt); break; case NPY_FR_fs: /* entire range is only +- 2.6 hours */ - out->hour = (int)extract_unit(&dt, 1000LL*1000*1000*1000*1000*60*60); + out->hour = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*1000*60*60); if (out->hour < 0) { out->year = 1969; out->month = 12; @@ -559,16 +559,16 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, out->hour += 24; assert(out->hour >= 0); } - out->min = (int)extract_unit(&dt, 1000LL*1000*1000*1000*1000*60); - out->sec = (int)extract_unit(&dt, 1000LL*1000*1000*1000*1000); - out->us = (int)extract_unit(&dt, 1000LL*1000*1000); - out->ps = (int)extract_unit(&dt, 1000LL); + out->min = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*1000*60); + out->sec = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*1000); + out->us = (int)extract_unit_64(&dt, 1000LL*1000*1000); + out->ps = (int)extract_unit_64(&dt, 1000LL); out->as = (int)(dt * 1000); break; case NPY_FR_as: /* entire range is only +- 9.2 seconds */ - out->sec = (int)extract_unit(&dt, 1000LL*1000*1000*1000*1000*1000); + out->sec = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000*1000*1000); if (out->sec < 0) { out->year = 1969; out->month = 12; @@ -578,8 +578,8 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, out->sec += 60; assert(out->sec >= 0); } - out->us = (int)extract_unit(&dt, 1000LL*1000*1000*1000); - out->ps = (int)extract_unit(&dt, 1000LL*1000); + out->us = (int)extract_unit_64(&dt, 1000LL*1000*1000*1000); + out->ps = (int)extract_unit_64(&dt, 1000LL*1000); out->as = (int)dt; break; @@ -600,8 +600,9 @@ convert_datetime_to_datetimestruct(PyArray_DatetimeMetaData *meta, * TO BE REMOVED - NOT USED INTERNALLY. */ NPY_NO_EXPORT void -PyArray_DatetimeToDatetimeStruct(npy_datetime val, NPY_DATETIMEUNIT fr, - npy_datetimestruct *result) +PyArray_DatetimeToDatetimeStruct( + npy_datetime NPY_UNUSED(val), NPY_DATETIMEUNIT NPY_UNUSED(fr), + npy_datetimestruct *result) { PyErr_SetString(PyExc_RuntimeError, "The NumPy PyArray_DatetimeToDatetimeStruct function has " @@ -621,8 +622,9 @@ PyArray_DatetimeToDatetimeStruct(npy_datetime val, NPY_DATETIMEUNIT fr, * TO BE REMOVED - NOT USED INTERNALLY. */ NPY_NO_EXPORT void -PyArray_TimedeltaToTimedeltaStruct(npy_timedelta val, NPY_DATETIMEUNIT fr, - npy_timedeltastruct *result) +PyArray_TimedeltaToTimedeltaStruct( + npy_timedelta NPY_UNUSED(val), NPY_DATETIMEUNIT NPY_UNUSED(fr), + npy_timedeltastruct *result) { PyErr_SetString(PyExc_RuntimeError, "The NumPy PyArray_TimedeltaToTimedeltaStruct function has " @@ -756,8 +758,8 @@ parse_datetime_extended_unit_from_string(char *str, Py_ssize_t len, bad_input: if (metastr != NULL) { PyErr_Format(PyExc_TypeError, - "Invalid datetime metadata string \"%s\" at position %d", - metastr, (int)(substr-metastr)); + "Invalid datetime metadata string \"%s\" at position %zd", + metastr, substr-metastr); } else { PyErr_Format(PyExc_TypeError, @@ -818,8 +820,8 @@ parse_datetime_metadata_from_metastr(char *metastr, Py_ssize_t len, bad_input: if (substr != metastr) { PyErr_Format(PyExc_TypeError, - "Invalid datetime metadata string \"%s\" at position %d", - metastr, (int)(substr-metastr)); + "Invalid datetime metadata string \"%s\" at position %zd", + metastr, substr - metastr); } else { PyErr_Format(PyExc_TypeError, @@ -1832,6 +1834,7 @@ convert_datetime_metadata_tuple_to_datetime_metadata(PyObject *tuple, return -1; } equal_one = PyObject_RichCompareBool(event, one, Py_EQ); + Py_DECREF(one); if (equal_one == -1) { return -1; } @@ -2012,20 +2015,8 @@ add_seconds_to_datetimestruct(npy_datetimestruct *dts, int seconds) int minutes; dts->sec += seconds; - if (dts->sec < 0) { - minutes = dts->sec / 60; - dts->sec = dts->sec % 60; - if (dts->sec < 0) { - --minutes; - dts->sec += 60; - } - add_minutes_to_datetimestruct(dts, minutes); - } - else if (dts->sec >= 60) { - minutes = dts->sec / 60; - dts->sec = dts->sec % 60; - add_minutes_to_datetimestruct(dts, minutes); - } + minutes = extract_unit_32(&dts->sec, 60); + add_minutes_to_datetimestruct(dts, minutes); } /* @@ -2037,28 +2028,13 @@ add_minutes_to_datetimestruct(npy_datetimestruct *dts, int minutes) { int isleap; - /* MINUTES */ dts->min += minutes; - while (dts->min < 0) { - dts->min += 60; - dts->hour--; - } - while (dts->min >= 60) { - dts->min -= 60; - dts->hour++; - } - /* HOURS */ - while (dts->hour < 0) { - dts->hour += 24; - dts->day--; - } - while (dts->hour >= 24) { - dts->hour -= 24; - dts->day++; - } + /* propagate invalid minutes into hour and day changes */ + dts->hour += extract_unit_32(&dts->min, 60); + dts->day += extract_unit_32(&dts->hour, 24); - /* DAYS */ + /* propagate invalid days into month and year changes */ if (dts->day < 1) { dts->month--; if (dts->month < 1) { @@ -2250,6 +2226,7 @@ convert_pydatetime_to_datetimestruct(PyObject *obj, npy_datetimestruct *out, if (DEPRECATE( "parsing timezone aware datetimes is deprecated; " "this will raise an error in the future") < 0) { + Py_DECREF(tmp); return -1; } @@ -2266,6 +2243,7 @@ convert_pydatetime_to_datetimestruct(PyObject *obj, npy_datetimestruct *out, * which contains the value we want. */ tmp = PyObject_CallMethod(offset, "total_seconds", ""); + Py_DECREF(offset); if (tmp == NULL) { return -1; } @@ -2295,15 +2273,15 @@ convert_pydatetime_to_datetimestruct(PyObject *obj, npy_datetimestruct *out, invalid_date: PyErr_Format(PyExc_ValueError, - "Invalid date (%d,%d,%d) when converting to NumPy datetime", - (int)out->year, (int)out->month, (int)out->day); + "Invalid date (%" NPY_INT64_FMT ",%" NPY_INT32_FMT ",%" NPY_INT32_FMT ") when converting to NumPy datetime", + out->year, out->month, out->day); return -1; invalid_time: PyErr_Format(PyExc_ValueError, - "Invalid time (%d,%d,%d,%d) when converting " + "Invalid time (%" NPY_INT32_FMT ",%" NPY_INT32_FMT ",%" NPY_INT32_FMT ",%" NPY_INT32_FMT ") when converting " "to NumPy datetime", - (int)out->hour, (int)out->min, (int)out->sec, (int)out->us); + out->hour, out->min, out->sec, out->us); return -1; } @@ -2883,7 +2861,6 @@ convert_datetime_to_pyobject(npy_datetime dt, PyArray_DatetimeMetaData *meta) NPY_NO_EXPORT PyObject * convert_timedelta_to_pyobject(npy_timedelta td, PyArray_DatetimeMetaData *meta) { - PyObject *ret = NULL; npy_timedelta value; int days = 0, seconds = 0, useconds = 0; @@ -2913,54 +2890,47 @@ convert_timedelta_to_pyobject(npy_timedelta td, PyArray_DatetimeMetaData *meta) /* Convert to days/seconds/useconds */ switch (meta->base) { case NPY_FR_W: - value *= 7; + days = value * 7; break; case NPY_FR_D: + days = value; break; case NPY_FR_h: - seconds = (int)((value % 24) * (60*60)); - value = value / 24; + days = extract_unit_64(&value, 24ULL); + seconds = value*60*60; break; case NPY_FR_m: - seconds = (int)(value % (24*60)) * 60; - value = value / (24*60); + days = extract_unit_64(&value, 60ULL*24); + seconds = value*60; break; case NPY_FR_s: - seconds = (int)(value % (24*60*60)); - value = value / (24*60*60); + days = extract_unit_64(&value, 60ULL*60*24); + seconds = value; break; case NPY_FR_ms: - useconds = (int)(value % 1000) * 1000; - value = value / 1000; - seconds = (int)(value % (24*60*60)); - value = value / (24*60*60); + days = extract_unit_64(&value, 1000ULL*60*60*24); + seconds = extract_unit_64(&value, 1000ULL); + useconds = value*1000; break; case NPY_FR_us: - useconds = (int)(value % (1000*1000)); - value = value / (1000*1000); - seconds = (int)(value % (24*60*60)); - value = value / (24*60*60); + days = extract_unit_64(&value, 1000ULL*1000*60*60*24); + seconds = extract_unit_64(&value, 1000ULL*1000); + useconds = value; break; default: + // unreachable, handled by the `if` above + assert(NPY_FALSE); break; } /* - * 'value' represents days, and seconds/useconds are filled. - * * If it would overflow the datetime.timedelta days, return a raw int */ - if (value < -999999999 || value > 999999999) { + if (days < -999999999 || days > 999999999) { return PyLong_FromLongLong(td); } else { - days = (int)value; - ret = PyDelta_FromDSU(days, seconds, useconds); - if (ret == NULL) { - return NULL; - } + return PyDelta_FromDSU(days, seconds, useconds); } - - return ret; } /* @@ -3125,7 +3095,7 @@ is_any_numpy_datetime_or_timedelta(PyObject *obj) */ NPY_NO_EXPORT int convert_pyobjects_to_datetimes(int count, - PyObject **objs, int *type_nums, + PyObject **objs, const int *type_nums, NPY_CASTING casting, npy_int64 *out_values, PyArray_DatetimeMetaData *inout_meta) @@ -3251,18 +3221,6 @@ NPY_NO_EXPORT PyArrayObject * datetime_arange(PyObject *start, PyObject *stop, PyObject *step, PyArray_Descr *dtype) { - PyArray_DatetimeMetaData meta; - /* - * Both datetime and timedelta are stored as int64, so they can - * share value variables. - */ - npy_int64 values[3]; - PyObject *objs[3]; - int type_nums[3]; - - npy_intp i, length; - PyArrayObject *ret; - npy_int64 *ret_data; /* * First normalize the input parameters so there is no Py_None, @@ -3295,6 +3253,8 @@ datetime_arange(PyObject *start, PyObject *stop, PyObject *step, /* Check if the units of the given dtype are generic, in which * case we use the code path that detects the units */ + int type_nums[3]; + PyArray_DatetimeMetaData meta; if (dtype != NULL) { PyArray_DatetimeMetaData *meta_tmp; @@ -3343,6 +3303,7 @@ datetime_arange(PyObject *start, PyObject *stop, PyObject *step, } /* Set up to convert the objects to a common datetime unit metadata */ + PyObject *objs[3]; objs[0] = start; objs[1] = stop; objs[2] = step; @@ -3363,11 +3324,22 @@ datetime_arange(PyObject *start, PyObject *stop, PyObject *step, type_nums[2] = NPY_TIMEDELTA; } - /* Convert all the arguments */ + /* Convert all the arguments + * + * Both datetime and timedelta are stored as int64, so they can + * share value variables. + */ + npy_int64 values[3]; if (convert_pyobjects_to_datetimes(3, objs, type_nums, NPY_SAME_KIND_CASTING, values, &meta) < 0) { return NULL; } + /* If no start was provided, default to 0 */ + if (start == NULL) { + /* enforced above */ + assert(type_nums[0] == NPY_TIMEDELTA); + values[0] = 0; + } /* If no step was provided, default to 1 */ if (step == NULL) { @@ -3392,6 +3364,7 @@ datetime_arange(PyObject *start, PyObject *stop, PyObject *step, } /* Calculate the array length */ + npy_intp length; if (values[2] > 0 && values[1] > values[0]) { length = (values[1] - values[0] + (values[2] - 1)) / values[2]; } @@ -3419,19 +3392,20 @@ datetime_arange(PyObject *start, PyObject *stop, PyObject *step, } /* Create the result array */ - ret = (PyArrayObject *)PyArray_NewFromDescr( - &PyArray_Type, dtype, 1, &length, NULL, - NULL, 0, NULL); + PyArrayObject *ret = (PyArrayObject *)PyArray_NewFromDescr( + &PyArray_Type, dtype, 1, &length, NULL, + NULL, 0, NULL); + if (ret == NULL) { return NULL; } if (length > 0) { /* Extract the data pointer */ - ret_data = (npy_int64 *)PyArray_DATA(ret); + npy_int64 *ret_data = (npy_int64 *)PyArray_DATA(ret); /* Create the timedeltas or datetimes */ - for (i = 0; i < length; ++i) { + for (npy_intp i = 0; i < length; ++i) { *ret_data = values[0]; values[0] += values[2]; ret_data++; diff --git a/numpy/core/src/multiarray/datetime_busday.c b/numpy/core/src/multiarray/datetime_busday.c index c04a6c125..cdeb65d0e 100644 --- a/numpy/core/src/multiarray/datetime_busday.c +++ b/numpy/core/src/multiarray/datetime_busday.c @@ -48,7 +48,7 @@ get_day_of_week(npy_datetime date) */ static int is_holiday(npy_datetime date, - npy_datetime *holidays_begin, npy_datetime *holidays_end) + npy_datetime *holidays_begin, const npy_datetime *holidays_end) { npy_datetime *trial; @@ -88,7 +88,7 @@ is_holiday(npy_datetime date, */ static npy_datetime * find_earliest_holiday_on_or_after(npy_datetime date, - npy_datetime *holidays_begin, npy_datetime *holidays_end) + npy_datetime *holidays_begin, const npy_datetime *holidays_end) { npy_datetime *trial; @@ -127,7 +127,7 @@ find_earliest_holiday_on_or_after(npy_datetime date, */ static npy_datetime * find_earliest_holiday_after(npy_datetime date, - npy_datetime *holidays_begin, npy_datetime *holidays_end) + npy_datetime *holidays_begin, const npy_datetime *holidays_end) { npy_datetime *trial; @@ -159,7 +159,7 @@ static int apply_business_day_roll(npy_datetime date, npy_datetime *out, int *out_day_of_week, NPY_BUSDAY_ROLL roll, - npy_bool *weekmask, + const npy_bool *weekmask, npy_datetime *holidays_begin, npy_datetime *holidays_end) { int day_of_week; @@ -361,7 +361,7 @@ apply_business_day_offset(npy_datetime date, npy_int64 offset, static int apply_business_day_count(npy_datetime date_begin, npy_datetime date_end, npy_int64 *out, - npy_bool *weekmask, int busdays_in_weekmask, + const npy_bool *weekmask, int busdays_in_weekmask, npy_datetime *holidays_begin, npy_datetime *holidays_end) { npy_int64 count, whole_weeks; @@ -722,7 +722,7 @@ finish: */ NPY_NO_EXPORT PyArrayObject * is_business_day(PyArrayObject *dates, PyArrayObject *out, - npy_bool *weekmask, int busdays_in_weekmask, + const npy_bool *weekmask, int busdays_in_weekmask, npy_datetime *holidays_begin, npy_datetime *holidays_end) { PyArray_DatetimeMetaData temp_meta; diff --git a/numpy/core/src/multiarray/datetime_strings.c b/numpy/core/src/multiarray/datetime_strings.c index 95b7bb3dc..dfc01494f 100644 --- a/numpy/core/src/multiarray/datetime_strings.c +++ b/numpy/core/src/multiarray/datetime_strings.c @@ -743,8 +743,8 @@ finish: parse_error: PyErr_Format(PyExc_ValueError, - "Error parsing datetime string \"%s\" at position %d", - str, (int)(substr-str)); + "Error parsing datetime string \"%s\" at position %zd", + str, substr - str); return -1; error: diff --git a/numpy/core/src/multiarray/descriptor.c b/numpy/core/src/multiarray/descriptor.c index cb4d7964e..d4e18e457 100644 --- a/numpy/core/src/multiarray/descriptor.c +++ b/numpy/core/src/multiarray/descriptor.c @@ -102,6 +102,7 @@ _arraydescr_from_dtype_attr(PyObject *obj, PyArray_Descr **newdescr) if (Py_EnterRecursiveCall( " while trying to convert the given data type from its " "`.dtype` attribute.") != 0) { + Py_DECREF(dtypedescr); return 1; } @@ -148,7 +149,7 @@ array_set_typeDict(PyObject *NPY_UNUSED(ignored), PyObject *args) arg == '|' || arg == '=') static int -_check_for_commastring(char *type, Py_ssize_t len) +_check_for_commastring(const char *type, Py_ssize_t len) { Py_ssize_t i; int sqbracket; @@ -1148,8 +1149,8 @@ _convert_from_dict(PyObject *obj, int align) } Py_DECREF(off); if (offset < 0) { - PyErr_Format(PyExc_ValueError, "offset %d cannot be negative", - (int)offset); + PyErr_Format(PyExc_ValueError, "offset %ld cannot be negative", + offset); Py_DECREF(tup); Py_DECREF(ind); goto fail; @@ -1163,10 +1164,10 @@ _convert_from_dict(PyObject *obj, int align) /* If align=True, enforce field alignment */ if (align && offset % newdescr->alignment != 0) { PyErr_Format(PyExc_ValueError, - "offset %d for NumPy dtype with fields is " + "offset %ld for NumPy dtype with fields is " "not divisible by the field alignment %d " "with align=True", - (int)offset, (int)newdescr->alignment); + offset, newdescr->alignment); ret = NPY_FAIL; } else if (offset + newdescr->elsize > totalsize) { @@ -1285,7 +1286,7 @@ _convert_from_dict(PyObject *obj, int align) PyErr_Format(PyExc_ValueError, "NumPy dtype descriptor requires %d bytes, " "cannot override to smaller itemsize of %d", - (int)new->elsize, (int)itemsize); + new->elsize, itemsize); Py_DECREF(new); goto fail; } @@ -1294,7 +1295,7 @@ _convert_from_dict(PyObject *obj, int align) PyErr_Format(PyExc_ValueError, "NumPy dtype descriptor requires alignment of %d bytes, " "which is not divisible into the specified itemsize %d", - (int)new->alignment, (int)itemsize); + new->alignment, itemsize); Py_DECREF(new); goto fail; } @@ -1384,7 +1385,6 @@ NPY_NO_EXPORT int PyArray_DescrConverter(PyObject *obj, PyArray_Descr **at) { int check_num = NPY_NOTYPE + 10; - PyObject *item; int elsize = 0; char endian = '='; @@ -1663,16 +1663,22 @@ finish: PyErr_Clear(); /* Now check to see if the object is registered in typeDict */ if (typeDict != NULL) { - item = PyDict_GetItem(typeDict, obj); + PyObject *item = NULL; #if defined(NPY_PY3K) - if (!item && PyBytes_Check(obj)) { + if (PyBytes_Check(obj)) { PyObject *tmp; tmp = PyUnicode_FromEncodedObject(obj, "ascii", "strict"); - if (tmp != NULL) { - item = PyDict_GetItem(typeDict, tmp); - Py_DECREF(tmp); + if (tmp == NULL) { + goto fail; } + item = PyDict_GetItem(typeDict, tmp); + Py_DECREF(tmp); + } + else { + item = PyDict_GetItem(typeDict, obj); } +#else + item = PyDict_GetItem(typeDict, obj); #endif if (item) { /* Check for a deprecated Numeric-style typecode */ @@ -3276,7 +3282,7 @@ arraydescr_richcompare(PyArray_Descr *self, PyObject *other, int cmp_op) } static int -descr_nonzero(PyObject *self) +descr_nonzero(PyObject *NPY_UNUSED(self)) { /* `bool(np.dtype(...)) == True` for all dtypes. Needed to override default * nonzero implementation, which checks if `len(object) > 0`. */ diff --git a/numpy/core/src/multiarray/getset.c b/numpy/core/src/multiarray/getset.c index 2c4969d23..6e5d480d0 100644 --- a/numpy/core/src/multiarray/getset.c +++ b/numpy/core/src/multiarray/getset.c @@ -190,7 +190,7 @@ array_strides_set(PyArrayObject *self, PyObject *obj) static PyObject * -array_priority_get(PyArrayObject *self) +array_priority_get(PyArrayObject *NPY_UNUSED(self)) { return PyFloat_FromDouble(NPY_PRIORITY); } diff --git a/numpy/core/src/multiarray/item_selection.c b/numpy/core/src/multiarray/item_selection.c index 01d9ecfb3..a6ac902d3 100644 --- a/numpy/core/src/multiarray/item_selection.c +++ b/numpy/core/src/multiarray/item_selection.c @@ -1336,7 +1336,11 @@ PyArray_ArgPartition(PyArrayObject *op, PyArrayObject *ktharray, int axis, PyArray_ArgSortFunc *argsort; PyObject *ret; - if (which < 0 || which >= NPY_NSELECTS) { + /* + * As a C-exported function, enum NPY_SELECTKIND loses its enum property + * Check the values to make sure they are in range + */ + if ((int)which < 0 || (int)which >= NPY_NSELECTS) { PyErr_SetString(PyExc_ValueError, "not a valid partition kind"); return NULL; @@ -2214,9 +2218,13 @@ PyArray_Nonzero(PyArrayObject *self) PyArrayObject *ret = NULL; PyObject *ret_tuple; npy_intp ret_dims[2]; - PyArray_NonzeroFunc *nonzero = PyArray_DESCR(self)->f->nonzero; + + PyArray_NonzeroFunc *nonzero; + PyArray_Descr *dtype; + npy_intp nonzero_count; npy_intp added_count = 0; + int needs_api; int is_bool; NpyIter *iter; @@ -2224,6 +2232,10 @@ PyArray_Nonzero(PyArrayObject *self) NpyIter_GetMultiIndexFunc *get_multi_index; char **dataptr; + dtype = PyArray_DESCR(self); + nonzero = dtype->f->nonzero; + needs_api = PyDataType_FLAGCHK(dtype, NPY_NEEDS_PYAPI); + /* Special case - nonzero(zero_d) is nonzero(atleast_1d(zero_d)) */ if (ndim == 0) { char const* msg; @@ -2247,6 +2259,7 @@ PyArray_Nonzero(PyArrayObject *self) static npy_intp const zero_dim_shape[1] = {1}; static npy_intp const zero_dim_strides[1] = {0}; + Py_INCREF(PyArray_DESCR(self)); /* array creation steals reference */ PyArrayObject *self_1d = (PyArrayObject *)PyArray_NewFromDescrAndBase( Py_TYPE(self), PyArray_DESCR(self), 1, zero_dim_shape, zero_dim_strides, PyArray_BYTES(self), @@ -2254,7 +2267,9 @@ PyArray_Nonzero(PyArrayObject *self) if (self_1d == NULL) { return NULL; } - return PyArray_Nonzero(self_1d); + ret_tuple = PyArray_Nonzero(self_1d); + Py_DECREF(self_1d); + return ret_tuple; } /* @@ -2291,7 +2306,9 @@ PyArray_Nonzero(PyArrayObject *self) goto finish; } - NPY_BEGIN_THREADS_THRESHOLDED(count); + if (!needs_api) { + NPY_BEGIN_THREADS_THRESHOLDED(count); + } /* avoid function call for bool */ if (is_bool) { @@ -2332,6 +2349,9 @@ PyArray_Nonzero(PyArrayObject *self) } *multi_index++ = j; } + if (needs_api && PyErr_Occurred()) { + break; + } data += stride; } } @@ -2372,6 +2392,8 @@ PyArray_Nonzero(PyArrayObject *self) Py_DECREF(ret); return NULL; } + + needs_api = NpyIter_IterationNeedsAPI(iter); NPY_BEGIN_THREADS_NDITER(iter); @@ -2398,6 +2420,9 @@ PyArray_Nonzero(PyArrayObject *self) get_multi_index(iter, multi_index); multi_index += ndim; } + if (needs_api && PyErr_Occurred()) { + break; + } } while(iternext(iter)); } @@ -2407,6 +2432,11 @@ PyArray_Nonzero(PyArrayObject *self) NpyIter_Deallocate(iter); finish: + if (PyErr_Occurred()) { + Py_DECREF(ret); + return NULL; + } + /* if executed `nonzero()` check for miscount due to side-effect */ if (!is_bool && added_count != nonzero_count) { PyErr_SetString(PyExc_RuntimeError, @@ -2449,7 +2479,7 @@ finish: * array of values, which must be of length PyArray_NDIM(self). */ NPY_NO_EXPORT PyObject * -PyArray_MultiIndexGetItem(PyArrayObject *self, npy_intp *multi_index) +PyArray_MultiIndexGetItem(PyArrayObject *self, const npy_intp *multi_index) { int idim, ndim = PyArray_NDIM(self); char *data = PyArray_DATA(self); @@ -2477,7 +2507,7 @@ PyArray_MultiIndexGetItem(PyArrayObject *self, npy_intp *multi_index) * Returns 0 on success, -1 on failure. */ NPY_NO_EXPORT int -PyArray_MultiIndexSetItem(PyArrayObject *self, npy_intp *multi_index, +PyArray_MultiIndexSetItem(PyArrayObject *self, const npy_intp *multi_index, PyObject *obj) { int idim, ndim = PyArray_NDIM(self); diff --git a/numpy/core/src/multiarray/item_selection.h b/numpy/core/src/multiarray/item_selection.h index 90bb5100d..2276b4db7 100644 --- a/numpy/core/src/multiarray/item_selection.h +++ b/numpy/core/src/multiarray/item_selection.h @@ -15,7 +15,7 @@ count_boolean_trues(int ndim, char *data, npy_intp *ashape, npy_intp *astrides); * array of values, which must be of length PyArray_NDIM(self). */ NPY_NO_EXPORT PyObject * -PyArray_MultiIndexGetItem(PyArrayObject *self, npy_intp *multi_index); +PyArray_MultiIndexGetItem(PyArrayObject *self, const npy_intp *multi_index); /* * Sets a single item in the array, based on a single multi-index @@ -24,7 +24,7 @@ PyArray_MultiIndexGetItem(PyArrayObject *self, npy_intp *multi_index); * Returns 0 on success, -1 on failure. */ NPY_NO_EXPORT int -PyArray_MultiIndexSetItem(PyArrayObject *self, npy_intp *multi_index, +PyArray_MultiIndexSetItem(PyArrayObject *self, const npy_intp *multi_index, PyObject *obj); #endif diff --git a/numpy/core/src/multiarray/iterators.c b/numpy/core/src/multiarray/iterators.c index 9da811f69..e66bb36aa 100644 --- a/numpy/core/src/multiarray/iterators.c +++ b/numpy/core/src/multiarray/iterators.c @@ -98,7 +98,7 @@ parse_index_entry(PyObject *op, npy_intp *step_size, /* get the dataptr from its current coordinates for simple iterator */ static char* -get_ptr_simple(PyArrayIterObject* iter, npy_intp *coordinates) +get_ptr_simple(PyArrayIterObject* iter, const npy_intp *coordinates) { npy_intp i; char *ret; @@ -116,10 +116,12 @@ get_ptr_simple(PyArrayIterObject* iter, npy_intp *coordinates) * This is common initialization code between PyArrayIterObject and * PyArrayNeighborhoodIterObject * - * Increase ao refcount + * Steals a reference to the array object which gets removed at deallocation, + * if the iterator is allocated statically and its dealloc not called, it + * can be thought of as borrowing the reference. */ -static PyObject * -array_iter_base_init(PyArrayIterObject *it, PyArrayObject *ao) +NPY_NO_EXPORT void +PyArray_RawIterBaseInit(PyArrayIterObject *it, PyArrayObject *ao) { int nd, i; @@ -131,7 +133,6 @@ array_iter_base_init(PyArrayIterObject *it, PyArrayObject *ao) else { it->contiguous = 0; } - Py_INCREF(ao); it->ao = ao; it->size = PyArray_SIZE(ao); it->nd_m1 = nd - 1; @@ -155,7 +156,7 @@ array_iter_base_init(PyArrayIterObject *it, PyArrayObject *ao) it->translate = &get_ptr_simple; PyArray_ITER_RESET(it); - return (PyObject *)it; + return; } static void @@ -170,6 +171,10 @@ array_iter_base_dealloc(PyArrayIterObject *it) NPY_NO_EXPORT PyObject * PyArray_IterNew(PyObject *obj) { + /* + * Note that internall PyArray_RawIterBaseInit may be called directly on a + * statically allocated PyArrayIterObject. + */ PyArrayIterObject *it; PyArrayObject *ao; @@ -186,7 +191,8 @@ PyArray_IterNew(PyObject *obj) return NULL; } - array_iter_base_init(it, ao); + Py_INCREF(ao); /* PyArray_RawIterBaseInit steals a reference */ + PyArray_RawIterBaseInit(it, ao); return (PyObject *)it; } @@ -390,6 +396,10 @@ arrayiter_next(PyArrayIterObject *it) static void arrayiter_dealloc(PyArrayIterObject *it) { + /* + * Note that it is possible to statically allocate a PyArrayIterObject, + * which does not call this function. + */ array_iter_base_dealloc(it); PyArray_free(it); } @@ -1398,6 +1408,7 @@ arraymultiter_new(PyTypeObject *NPY_UNUSED(subtype), PyObject *args, } n = PySequence_Fast_GET_SIZE(fast_seq); if (n > NPY_MAXARGS) { + Py_DECREF(fast_seq); return multiiter_wrong_number_of_args(); } ret = multiiter_new_impl(n, PySequence_Fast_ITEMS(fast_seq)); @@ -1654,7 +1665,7 @@ static char* _set_constant(PyArrayNeighborhoodIterObject* iter, /* set the dataptr from its current coordinates */ static char* -get_ptr_constant(PyArrayIterObject* _iter, npy_intp *coordinates) +get_ptr_constant(PyArrayIterObject* _iter, const npy_intp *coordinates) { int i; npy_intp bd, _coordinates[NPY_MAXDIMS]; @@ -1709,7 +1720,7 @@ __npy_pos_remainder(npy_intp i, npy_intp n) /* set the dataptr from its current coordinates */ static char* -get_ptr_mirror(PyArrayIterObject* _iter, npy_intp *coordinates) +get_ptr_mirror(PyArrayIterObject* _iter, const npy_intp *coordinates) { int i; npy_intp bd, _coordinates[NPY_MAXDIMS], lb; @@ -1743,7 +1754,7 @@ __npy_euclidean_division(npy_intp i, npy_intp n) _coordinates[c] = lb + __npy_euclidean_division(bd, p->limits_sizes[c]); static char* -get_ptr_circular(PyArrayIterObject* _iter, npy_intp *coordinates) +get_ptr_circular(PyArrayIterObject* _iter, const npy_intp *coordinates) { int i; npy_intp bd, _coordinates[NPY_MAXDIMS], lb; @@ -1765,7 +1776,7 @@ get_ptr_circular(PyArrayIterObject* _iter, npy_intp *coordinates) * A Neighborhood Iterator object. */ NPY_NO_EXPORT PyObject* -PyArray_NeighborhoodIterNew(PyArrayIterObject *x, npy_intp *bounds, +PyArray_NeighborhoodIterNew(PyArrayIterObject *x, const npy_intp *bounds, int mode, PyArrayObject* fill) { int i; @@ -1777,7 +1788,8 @@ PyArray_NeighborhoodIterNew(PyArrayIterObject *x, npy_intp *bounds, } PyObject_Init((PyObject *)ret, &PyArrayNeighborhoodIter_Type); - array_iter_base_init((PyArrayIterObject*)ret, x->ao); + Py_INCREF(x->ao); /* PyArray_RawIterBaseInit steals a reference */ + PyArray_RawIterBaseInit((PyArrayIterObject*)ret, x->ao); Py_INCREF(x); ret->_internal_iter = x; diff --git a/numpy/core/src/multiarray/iterators.h b/numpy/core/src/multiarray/iterators.h index 376dc154a..d942f45b8 100644 --- a/numpy/core/src/multiarray/iterators.h +++ b/numpy/core/src/multiarray/iterators.h @@ -7,4 +7,7 @@ NPY_NO_EXPORT PyObject NPY_NO_EXPORT int iter_ass_subscript(PyArrayIterObject *, PyObject *, PyObject *); +NPY_NO_EXPORT void +PyArray_RawIterBaseInit(PyArrayIterObject *it, PyArrayObject *ao); + #endif diff --git a/numpy/core/src/multiarray/mapping.c b/numpy/core/src/multiarray/mapping.c index cc628e47e..8dcd28c84 100644 --- a/numpy/core/src/multiarray/mapping.c +++ b/numpy/core/src/multiarray/mapping.c @@ -176,7 +176,7 @@ unpack_tuple(PyTupleObject *index, PyObject **result, npy_intp result_n) /* Unpack a single scalar index, taking a new reference to match unpack_tuple */ static NPY_INLINE npy_intp -unpack_scalar(PyObject *index, PyObject **result, npy_intp result_n) +unpack_scalar(PyObject *index, PyObject **result, npy_intp NPY_UNUSED(result_n)) { Py_INCREF(index); result[0] = index; @@ -1198,9 +1198,9 @@ array_assign_boolean_subscript(PyArrayObject *self, if (size != PyArray_DIMS(v)[0]) { PyErr_Format(PyExc_ValueError, "NumPy boolean array indexing assignment " - "cannot assign %d input values to " - "the %d output values where the mask is true", - (int)PyArray_DIMS(v)[0], (int)size); + "cannot assign %" NPY_INTP_FMT " input values to " + "the %" NPY_INTP_FMT " output values where the mask is true", + PyArray_DIMS(v)[0], size); return -1; } v_stride = PyArray_STRIDES(v)[0]; @@ -2516,6 +2516,7 @@ PyArray_MapIterCheckIndices(PyArrayMapIterObject *mit) indval = *((npy_intp*)data); if (check_and_adjust_index(&indval, outer_dim, outer_axis, _save) < 0) { + Py_DECREF(intp_type); return -1; } data += stride; @@ -2616,7 +2617,8 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, PyArrayObject *original_extra_op = extra_op; PyArrayObject *index_arrays[NPY_MAXDIMS]; - PyArray_Descr *dtypes[NPY_MAXDIMS]; + PyArray_Descr *intp_descr; + PyArray_Descr *dtypes[NPY_MAXDIMS]; /* borrowed references */ npy_uint32 op_flags[NPY_MAXDIMS]; npy_uint32 outer_flags; @@ -2629,9 +2631,15 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, int nops; int uses_subspace; + intp_descr = PyArray_DescrFromType(NPY_INTP); + if (intp_descr == NULL) { + return NULL; + } + /* create new MapIter object */ mit = (PyArrayMapIterObject *)PyArray_malloc(sizeof(PyArrayMapIterObject)); if (mit == NULL) { + Py_DECREF(intp_descr); return NULL; } /* set all attributes of mapiter to zero */ @@ -2661,6 +2669,7 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, mit->nd_fancy = fancy_ndim; if (mapiter_fill_info(mit, indices, index_num, arr) < 0) { Py_DECREF(mit); + Py_DECREF(intp_descr); return NULL; } @@ -2670,7 +2679,7 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, for (i=0; i < index_num; i++) { if (indices[i].type & HAS_FANCY) { index_arrays[mit->numiter] = (PyArrayObject *)indices[i].object; - dtypes[mit->numiter] = PyArray_DescrFromType(NPY_INTP); + dtypes[mit->numiter] = intp_descr; op_flags[mit->numiter] = (NPY_ITER_NBO | NPY_ITER_ALIGNED | @@ -2693,9 +2702,10 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, PyArray_DescrFromType(NPY_INTP), 0); if (index_arrays[0] == NULL) { Py_DECREF(mit); + Py_DECREF(intp_descr); return NULL; } - dtypes[0] = PyArray_DescrFromType(NPY_INTP); + dtypes[0] = intp_descr; op_flags[0] = NPY_ITER_NBO | NPY_ITER_ALIGNED | NPY_ITER_READONLY; mit->fancy_dims[0] = 1; @@ -2925,7 +2935,6 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, nops += 1; index_arrays[mit->numiter] = extra_op; - Py_INCREF(extra_op_dtype); dtypes[mit->numiter] = extra_op_dtype; op_flags[mit->numiter] = (extra_op_flags | NPY_ITER_ALLOCATE | @@ -2951,9 +2960,6 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, } /* NpyIter cleanup and information: */ - for (i=0; i < nops; i++) { - Py_DECREF(dtypes[i]); - } if (dummy_array) { Py_DECREF(index_arrays[0]); } @@ -3039,6 +3045,7 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, /* Can now return early if no subspace is being used */ if (!uses_subspace) { Py_XDECREF(extra_op); + Py_DECREF(intp_descr); return (PyObject *)mit; } @@ -3108,6 +3115,7 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, } Py_XDECREF(extra_op); + Py_DECREF(intp_descr); return (PyObject *)mit; fail: @@ -3176,6 +3184,7 @@ PyArray_MapIterNew(npy_index_info *indices , int index_num, int index_type, finish: Py_XDECREF(extra_op); + Py_DECREF(intp_descr); Py_DECREF(mit); return NULL; } diff --git a/numpy/core/src/multiarray/methods.c b/numpy/core/src/multiarray/methods.c index d458638ae..e5845f2f6 100644 --- a/numpy/core/src/multiarray/methods.c +++ b/numpy/core/src/multiarray/methods.c @@ -365,6 +365,7 @@ PyArray_GetField(PyArrayObject *self, PyArray_Descr *typed, int offset) npy_cache_import("numpy.core._internal", "_getfield_is_safe", &checkfunc); if (checkfunc == NULL) { + Py_DECREF(typed); return NULL; } @@ -372,6 +373,7 @@ PyArray_GetField(PyArrayObject *self, PyArray_Descr *typed, int offset) safe = PyObject_CallFunction(checkfunc, "OOi", PyArray_DESCR(self), typed, offset); if (safe == NULL) { + Py_DECREF(typed); return NULL; } Py_DECREF(safe); @@ -382,14 +384,17 @@ PyArray_GetField(PyArrayObject *self, PyArray_Descr *typed, int offset) /* check that values are valid */ if (typed_elsize > self_elsize) { PyErr_SetString(PyExc_ValueError, "new type is larger than original type"); + Py_DECREF(typed); return NULL; } if (offset < 0) { PyErr_SetString(PyExc_ValueError, "offset is negative"); + Py_DECREF(typed); return NULL; } if (offset > self_elsize - typed_elsize) { PyErr_SetString(PyExc_ValueError, "new type plus offset is larger than original type"); + Py_DECREF(typed); return NULL; } @@ -434,6 +439,7 @@ PyArray_SetField(PyArrayObject *self, PyArray_Descr *dtype, int retval = 0; if (PyArray_FailUnlessWriteable(self, "assignment destination") < 0) { + Py_DECREF(dtype); return -1; } @@ -583,14 +589,13 @@ array_tofile(PyArrayObject *self, PyObject *args, PyObject *kwds) return NULL; } if (PyBytes_Check(file) || PyUnicode_Check(file)) { - file = npy_PyFile_OpenFile(file, "wb"); + Py_SETREF(file, npy_PyFile_OpenFile(file, "wb")); if (file == NULL) { return NULL; } own = 1; } else { - Py_INCREF(file); own = 0; } @@ -1046,7 +1051,7 @@ any_array_ufunc_overrides(PyObject *args, PyObject *kwds) NPY_NO_EXPORT PyObject * -array_ufunc(PyArrayObject *self, PyObject *args, PyObject *kwds) +array_ufunc(PyArrayObject *NPY_UNUSED(self), PyObject *args, PyObject *kwds) { PyObject *ufunc, *method_name, *normal_args, *ufunc_method; PyObject *result = NULL; @@ -1095,7 +1100,7 @@ cleanup: } static PyObject * -array_function(PyArrayObject *self, PyObject *c_args, PyObject *c_kwds) +array_function(PyArrayObject *NPY_UNUSED(self), PyObject *c_args, PyObject *c_kwds) { PyObject *func, *types, *args, *kwargs, *result; static char *kwlist[] = {"func", "types", "args", "kwargs", NULL}; @@ -1174,7 +1179,7 @@ array_resize(PyArrayObject *self, PyObject *args, PyObject *kwds) return NULL; } - ret = PyArray_Resize(self, &newshape, refcheck, NPY_CORDER); + ret = PyArray_Resize(self, &newshape, refcheck, NPY_ANYORDER); npy_free_cache_dim_obj(newshape); if (ret == NULL) { return NULL; @@ -1727,7 +1732,7 @@ array_reduce(PyArrayObject *self, PyObject *NPY_UNUSED(args)) } static PyObject * -array_reduce_ex_regular(PyArrayObject *self, int protocol) +array_reduce_ex_regular(PyArrayObject *self, int NPY_UNUSED(protocol)) { PyObject *subclass_array_reduce = NULL; PyObject *ret; @@ -1856,7 +1861,7 @@ array_reduce_ex(PyArrayObject *self, PyObject *args) PyDataType_FLAGCHK(descr, NPY_ITEM_HASOBJECT) || (PyType_IsSubtype(((PyObject*)self)->ob_type, &PyArray_Type) && ((PyObject*)self)->ob_type != &PyArray_Type) || - PyDataType_ISUNSIZED(descr)) { + descr->elsize == 0) { /* The PickleBuffer class from version 5 of the pickle protocol * can only be used for arrays backed by a contiguous data buffer. * For all other cases we fallback to the generic array_reduce @@ -2035,6 +2040,7 @@ array_setstate(PyArrayObject *self, PyObject *args) #endif npy_intp num = PyArray_NBYTES(self); if (num == 0) { + Py_DECREF(rawdata); Py_RETURN_NONE; } fa->data = PyDataMem_NEW(num); @@ -2385,7 +2391,6 @@ array_clip(PyArrayObject *self, PyObject *args, PyObject *kwds) static PyObject * array_conjugate(PyArrayObject *self, PyObject *args) { - PyArrayObject *out = NULL; if (!PyArg_ParseTuple(args, "|O&:conjugate", PyArray_OutputConverter, diff --git a/numpy/core/src/multiarray/methods.h b/numpy/core/src/multiarray/methods.h index b96a3c8a8..7a9a24a00 100644 --- a/numpy/core/src/multiarray/methods.h +++ b/numpy/core/src/multiarray/methods.h @@ -8,7 +8,10 @@ extern NPY_NO_EXPORT PyMethodDef array_methods[]; NPY_NO_EXPORT const char * npy_casting_to_string(NPY_CASTING casting); -/* Pathlib support */ +/* + * Pathlib support, takes a borrowed reference and returns a new one. + * The new object may be the same as the old. + */ static inline PyObject * NpyPath_PathlikeToFspath(PyObject *file) { @@ -24,6 +27,7 @@ NpyPath_PathlikeToFspath(PyObject *file) } if (!PyObject_IsInstance(file, os_PathLike)) { + Py_INCREF(file); return file; } return PyObject_CallFunctionObjArgs(os_fspath, file, NULL); diff --git a/numpy/core/src/multiarray/multiarraymodule.c b/numpy/core/src/multiarray/multiarraymodule.c index 084a5dd46..9169814c2 100644 --- a/numpy/core/src/multiarray/multiarraymodule.c +++ b/numpy/core/src/multiarray/multiarraymodule.c @@ -118,6 +118,9 @@ PyArray_GetPriority(PyObject *obj, double default_) ret = PyArray_LookupSpecial_OnInstance(obj, "__array_priority__"); if (ret == NULL) { + if (PyErr_Occurred()) { + PyErr_Clear(); /* TODO[gh-14801]: propagate crashes during attribute access? */ + } return default_; } @@ -286,45 +289,26 @@ PyArray_AsCArray(PyObject **op, void *ptr, npy_intp *dims, int nd, * Convert to a 1D C-array */ NPY_NO_EXPORT int -PyArray_As1D(PyObject **op, char **ptr, int *d1, int typecode) +PyArray_As1D(PyObject **NPY_UNUSED(op), char **NPY_UNUSED(ptr), + int *NPY_UNUSED(d1), int NPY_UNUSED(typecode)) { - npy_intp newd1; - PyArray_Descr *descr; - static const char msg[] = "PyArray_As1D: use PyArray_AsCArray."; - /* 2008-07-14, 1.5 */ - if (DEPRECATE(msg) < 0) { - return -1; - } - descr = PyArray_DescrFromType(typecode); - if (PyArray_AsCArray(op, (void *)ptr, &newd1, 1, descr) == -1) { - return -1; - } - *d1 = (int) newd1; - return 0; + PyErr_SetString(PyExc_NotImplementedError, + "PyArray_As1D: use PyArray_AsCArray."); + return -1; } /*NUMPY_API * Convert to a 2D C-array */ NPY_NO_EXPORT int -PyArray_As2D(PyObject **op, char ***ptr, int *d1, int *d2, int typecode) +PyArray_As2D(PyObject **NPY_UNUSED(op), char ***NPY_UNUSED(ptr), + int *NPY_UNUSED(d1), int *NPY_UNUSED(d2), int NPY_UNUSED(typecode)) { - npy_intp newdims[2]; - PyArray_Descr *descr; - static const char msg[] = "PyArray_As1D: use PyArray_AsCArray."; - /* 2008-07-14, 1.5 */ - if (DEPRECATE(msg) < 0) { - return -1; - } - descr = PyArray_DescrFromType(typecode); - if (PyArray_AsCArray(op, (void *)ptr, newdims, 2, descr) == -1) { - return -1; - } - *d1 = (int ) newdims[0]; - *d2 = (int ) newdims[1]; - return 0; + PyErr_SetString(PyExc_NotImplementedError, + "PyArray_As2D: use PyArray_AsCArray."); + return -1; } /* End Deprecated */ @@ -1131,6 +1115,14 @@ _pyarray_correlate(PyArrayObject *ap1, PyArrayObject *ap2, int typenum, n1 = PyArray_DIMS(ap1)[0]; n2 = PyArray_DIMS(ap2)[0]; + if (n1 == 0) { + PyErr_SetString(PyExc_ValueError, "first array argument cannot be empty"); + return NULL; + } + if (n2 == 0) { + PyErr_SetString(PyExc_ValueError, "second array argument cannot be empty"); + return NULL; + } if (n1 < n2) { ret = ap1; ap1 = ap2; @@ -2082,25 +2074,28 @@ array_fromfile(PyObject *NPY_UNUSED(ignored), PyObject *args, PyObject *keywds) if (file == NULL) { return NULL; } - + if (offset != 0 && strcmp(sep, "") != 0) { PyErr_SetString(PyExc_TypeError, "'offset' argument only permitted for binary files"); + Py_XDECREF(type); + Py_DECREF(file); return NULL; } if (PyString_Check(file) || PyUnicode_Check(file)) { - file = npy_PyFile_OpenFile(file, "rb"); + Py_SETREF(file, npy_PyFile_OpenFile(file, "rb")); if (file == NULL) { + Py_XDECREF(type); return NULL; } own = 1; } else { - Py_INCREF(file); own = 0; } fp = npy_PyFile_Dup2(file, "rb", &orig_pos); if (fp == NULL) { Py_DECREF(file); + Py_XDECREF(type); return NULL; } if (npy_fseek(fp, offset, SEEK_CUR) != 0) { @@ -3281,7 +3276,7 @@ array_datetime_data(PyObject *NPY_UNUSED(dummy), PyObject *args) } meta = get_datetime_metadata_from_dtype(dtype); - Py_DECREF(dtype); + Py_DECREF(dtype); if (meta == NULL) { return NULL; } @@ -3798,7 +3793,7 @@ _vec_string_no_args(PyArrayObject* char_array, } static PyObject * -_vec_string(PyObject *NPY_UNUSED(dummy), PyObject *args, PyObject *kwds) +_vec_string(PyObject *NPY_UNUSED(dummy), PyObject *args, PyObject *NPY_UNUSED(kwds)) { PyArrayObject* char_array = NULL; PyArray_Descr *type; @@ -3824,9 +3819,11 @@ _vec_string(PyObject *NPY_UNUSED(dummy), PyObject *args, PyObject *kwds) else { PyErr_SetString(PyExc_TypeError, "string operation on non-string array"); + Py_DECREF(type); goto err; } if (method == NULL) { + Py_DECREF(type); goto err; } diff --git a/numpy/core/src/multiarray/nditer_api.c b/numpy/core/src/multiarray/nditer_api.c index db0bfcece..e7fe0fa50 100644 --- a/numpy/core/src/multiarray/nditer_api.c +++ b/numpy/core/src/multiarray/nditer_api.c @@ -371,8 +371,8 @@ NpyIter_ResetToIterIndexRange(NpyIter *iter, } if (errmsg == NULL) { PyErr_Format(PyExc_ValueError, - "Out-of-bounds range [%d, %d) passed to " - "ResetToIterIndexRange", (int)istart, (int)iend); + "Out-of-bounds range [%" NPY_INTP_FMT ", %" NPY_INTP_FMT ") passed to " + "ResetToIterIndexRange", istart, iend); } else { *errmsg = "Out-of-bounds range passed to ResetToIterIndexRange"; @@ -382,8 +382,8 @@ NpyIter_ResetToIterIndexRange(NpyIter *iter, else if (iend < istart) { if (errmsg == NULL) { PyErr_Format(PyExc_ValueError, - "Invalid range [%d, %d) passed to ResetToIterIndexRange", - (int)istart, (int)iend); + "Invalid range [%" NPY_INTP_FMT ", %" NPY_INTP_FMT ") passed to ResetToIterIndexRange", + istart, iend); } else { *errmsg = "Invalid range passed to ResetToIterIndexRange"; @@ -1429,8 +1429,8 @@ NpyIter_DebugPrint(NpyIter *iter) printf("REUSE_REDUCE_LOOPS "); printf("\n"); - printf("| NDim: %d\n", (int)ndim); - printf("| NOp: %d\n", (int)nop); + printf("| NDim: %d\n", ndim); + printf("| NOp: %d\n", nop); if (NIT_MASKOP(iter) >= 0) { printf("| MaskOp: %d\n", (int)NIT_MASKOP(iter)); } diff --git a/numpy/core/src/multiarray/nditer_constr.c b/numpy/core/src/multiarray/nditer_constr.c index 3b3635afe..5e770338d 100644 --- a/numpy/core/src/multiarray/nditer_constr.c +++ b/numpy/core/src/multiarray/nditer_constr.c @@ -24,7 +24,7 @@ static int npyiter_check_global_flags(npy_uint32 flags, npy_uint32* itflags); static int npyiter_check_op_axes(int nop, int oa_ndim, int **op_axes, - npy_intp *itershape); + const npy_intp *itershape); static int npyiter_calculate_ndim(int nop, PyArrayObject **op_in, int oa_ndim); @@ -55,7 +55,7 @@ npyiter_check_casting(int nop, PyArrayObject **op, static int npyiter_fill_axisdata(NpyIter *iter, npy_uint32 flags, npyiter_opitflags *op_itflags, char **op_dataptr, - npy_uint32 *op_flags, int **op_axes, + const npy_uint32 *op_flags, int **op_axes, npy_intp *itershape); static void npyiter_replace_axisdata(NpyIter *iter, int iop, @@ -74,23 +74,23 @@ static void npyiter_find_best_axis_ordering(NpyIter *iter); static PyArray_Descr * npyiter_get_common_dtype(int nop, PyArrayObject **op, - npyiter_opitflags *op_itflags, PyArray_Descr **op_dtype, + const npyiter_opitflags *op_itflags, PyArray_Descr **op_dtype, PyArray_Descr **op_request_dtypes, int only_inputs); static PyArrayObject * npyiter_new_temp_array(NpyIter *iter, PyTypeObject *subtype, npy_uint32 flags, npyiter_opitflags *op_itflags, int op_ndim, npy_intp *shape, - PyArray_Descr *op_dtype, int *op_axes); + PyArray_Descr *op_dtype, const int *op_axes); static int npyiter_allocate_arrays(NpyIter *iter, npy_uint32 flags, PyArray_Descr **op_dtype, PyTypeObject *subtype, - npy_uint32 *op_flags, npyiter_opitflags *op_itflags, + const npy_uint32 *op_flags, npyiter_opitflags *op_itflags, int **op_axes); static void npyiter_get_priority_subtype(int nop, PyArrayObject **op, - npyiter_opitflags *op_itflags, + const npyiter_opitflags *op_itflags, double *subtype_priority, PyTypeObject **subtype); static int npyiter_allocate_transfer_functions(NpyIter *iter); @@ -154,7 +154,7 @@ NpyIter_AdvancedNew(int nop, PyArrayObject **op_in, npy_uint32 flags, if (nop > NPY_MAXARGS) { PyErr_Format(PyExc_ValueError, "Cannot construct an iterator with more than %d operands " - "(%d were requested)", (int)NPY_MAXARGS, (int)nop); + "(%d were requested)", NPY_MAXARGS, nop); return NULL; } @@ -787,7 +787,7 @@ npyiter_check_global_flags(npy_uint32 flags, npy_uint32* itflags) static int npyiter_check_op_axes(int nop, int oa_ndim, int **op_axes, - npy_intp *itershape) + const npy_intp *itershape) { char axes_dupcheck[NPY_MAXDIMS]; int iop, idim; @@ -810,7 +810,7 @@ npyiter_check_op_axes(int nop, int oa_ndim, int **op_axes, PyErr_Format(PyExc_ValueError, "Cannot construct an iterator with more than %d dimensions " "(%d were requested for op_axes)", - (int)NPY_MAXDIMS, oa_ndim); + NPY_MAXDIMS, oa_ndim); return 0; } if (op_axes == NULL) { @@ -826,14 +826,14 @@ npyiter_check_op_axes(int nop, int oa_ndim, int **op_axes, if (axes != NULL) { memset(axes_dupcheck, 0, NPY_MAXDIMS); for (idim = 0; idim < oa_ndim; ++idim) { - npy_intp i = axes[idim]; + int i = axes[idim]; if (i >= 0) { if (i >= NPY_MAXDIMS) { PyErr_Format(PyExc_ValueError, "The 'op_axes' provided to the iterator " "constructor for operand %d " "contained invalid " - "values %d", (int)iop, (int)i); + "values %d", iop, i); return 0; } else if (axes_dupcheck[i] == 1) { @@ -841,7 +841,7 @@ npyiter_check_op_axes(int nop, int oa_ndim, int **op_axes, "The 'op_axes' provided to the iterator " "constructor for operand %d " "contained duplicate " - "value %d", (int)iop, (int)i); + "value %d", iop, i); return 0; } else { @@ -1311,7 +1311,7 @@ npyiter_check_casting(int nop, PyArrayObject **op, PyObject *errmsg; errmsg = PyUString_FromFormat( "Iterator operand %d dtype could not be cast from ", - (int)iop); + iop); PyUString_ConcatAndDel(&errmsg, PyObject_Repr((PyObject *)PyArray_DESCR(op[iop]))); PyUString_ConcatAndDel(&errmsg, @@ -1342,7 +1342,7 @@ npyiter_check_casting(int nop, PyArrayObject **op, PyUString_ConcatAndDel(&errmsg, PyUString_FromFormat(", the operand %d dtype, " "according to the rule %s", - (int)iop, + iop, npyiter_casting_to_string(casting))); PyErr_SetObject(PyExc_TypeError, errmsg); Py_DECREF(errmsg); @@ -1423,7 +1423,7 @@ check_mask_for_writemasked_reduction(NpyIter *iter, int iop) static int npyiter_fill_axisdata(NpyIter *iter, npy_uint32 flags, npyiter_opitflags *op_itflags, char **op_dataptr, - npy_uint32 *op_flags, int **op_axes, + const npy_uint32 *op_flags, int **op_axes, npy_intp *itershape) { npy_uint32 itflags = NIT_ITFLAGS(iter); @@ -1500,8 +1500,8 @@ npyiter_fill_axisdata(NpyIter *iter, npy_uint32 flags, npyiter_opitflags *op_itf "Iterator input op_axes[%d][%d] (==%d) " "is not a valid axis of op[%d], which " "has %d dimensions ", - (int)iop, (int)(ndim-idim-1), (int)i, - (int)iop, (int)ondim); + iop, (ndim-idim-1), i, + iop, ondim); return 0; } } @@ -2409,7 +2409,7 @@ npyiter_find_best_axis_ordering(NpyIter *iter) */ static PyArray_Descr * npyiter_get_common_dtype(int nop, PyArrayObject **op, - npyiter_opitflags *op_itflags, PyArray_Descr **op_dtype, + const npyiter_opitflags *op_itflags, PyArray_Descr **op_dtype, PyArray_Descr **op_request_dtypes, int only_inputs) { @@ -2477,7 +2477,7 @@ static PyArrayObject * npyiter_new_temp_array(NpyIter *iter, PyTypeObject *subtype, npy_uint32 flags, npyiter_opitflags *op_itflags, int op_ndim, npy_intp *shape, - PyArray_Descr *op_dtype, int *op_axes) + PyArray_Descr *op_dtype, const int *op_axes) { npy_uint32 itflags = NIT_ITFLAGS(iter); int idim, ndim = NIT_NDIM(iter); @@ -2706,7 +2706,7 @@ static int npyiter_allocate_arrays(NpyIter *iter, npy_uint32 flags, PyArray_Descr **op_dtype, PyTypeObject *subtype, - npy_uint32 *op_flags, npyiter_opitflags *op_itflags, + const npy_uint32 *op_flags, npyiter_opitflags *op_itflags, int **op_axes) { npy_uint32 itflags = NIT_ITFLAGS(iter); @@ -3109,7 +3109,7 @@ npyiter_allocate_arrays(NpyIter *iter, */ static void npyiter_get_priority_subtype(int nop, PyArrayObject **op, - npyiter_opitflags *op_itflags, + const npyiter_opitflags *op_itflags, double *subtype_priority, PyTypeObject **subtype) { diff --git a/numpy/core/src/multiarray/nditer_pywrap.c b/numpy/core/src/multiarray/nditer_pywrap.c index 25d21025e..246f9d382 100644 --- a/numpy/core/src/multiarray/nditer_pywrap.c +++ b/numpy/core/src/multiarray/nditer_pywrap.c @@ -82,7 +82,8 @@ static int npyiter_cache_values(NewNpyArrayIterObject *self) } static PyObject * -npyiter_new(PyTypeObject *subtype, PyObject *args, PyObject *kwds) +npyiter_new(PyTypeObject *subtype, PyObject *NPY_UNUSED(args), + PyObject *NPY_UNUSED(kwds)) { NewNpyArrayIterObject *self; @@ -572,6 +573,7 @@ npyiter_convert_op_axes(PyObject *op_axes_in, int nop, if (*oa_ndim > NPY_MAXDIMS) { PyErr_SetString(PyExc_ValueError, "Too many dimensions in op_axes"); + Py_DECREF(a); return 0; } } @@ -602,8 +604,8 @@ npyiter_convert_op_axes(PyObject *op_axes_in, int nop, } Py_DECREF(v); } - Py_DECREF(a); } + Py_DECREF(a); } if (*oa_ndim == -1) { @@ -2014,7 +2016,7 @@ npyiter_seq_item(NewNpyArrayIterObject *self, Py_ssize_t i) if (i < 0 || i >= nop) { PyErr_Format(PyExc_IndexError, - "Iterator operand index %d is out of bounds", (int)i_orig); + "Iterator operand index %zd is out of bounds", i_orig); return NULL; } @@ -2028,7 +2030,7 @@ npyiter_seq_item(NewNpyArrayIterObject *self, Py_ssize_t i) */ if (!self->readflags[i]) { PyErr_Format(PyExc_RuntimeError, - "Iterator operand %d is write-only", (int)i); + "Iterator operand %zd is write-only", i); return NULL; } #endif @@ -2145,12 +2147,12 @@ npyiter_seq_ass_item(NewNpyArrayIterObject *self, Py_ssize_t i, PyObject *v) if (i < 0 || i >= nop) { PyErr_Format(PyExc_IndexError, - "Iterator operand index %d is out of bounds", (int)i_orig); + "Iterator operand index %zd is out of bounds", i_orig); return -1; } if (!self->writeflags[i]) { PyErr_Format(PyExc_RuntimeError, - "Iterator operand %d is not writeable", (int)i_orig); + "Iterator operand %zd is not writeable", i_orig); return -1; } @@ -2364,7 +2366,7 @@ npyiter_close(NewNpyArrayIterObject *self) } static PyObject * -npyiter_exit(NewNpyArrayIterObject *self, PyObject *args) +npyiter_exit(NewNpyArrayIterObject *self, PyObject *NPY_UNUSED(args)) { /* even if called via exception handling, writeback any data */ return npyiter_close(self); diff --git a/numpy/core/src/multiarray/number.c b/numpy/core/src/multiarray/number.c index 0ceb994ef..dabc866ff 100644 --- a/numpy/core/src/multiarray/number.c +++ b/numpy/core/src/multiarray/number.c @@ -391,7 +391,8 @@ array_matrix_multiply(PyArrayObject *m1, PyObject *m2) } static PyObject * -array_inplace_matrix_multiply(PyArrayObject *m1, PyObject *m2) +array_inplace_matrix_multiply( + PyArrayObject *NPY_UNUSED(m1), PyObject *NPY_UNUSED(m2)) { PyErr_SetString(PyExc_TypeError, "In-place matrix multiplication is not (yet) supported. " diff --git a/numpy/core/src/multiarray/refcount.c b/numpy/core/src/multiarray/refcount.c index b8230c81a..6033929d9 100644 --- a/numpy/core/src/multiarray/refcount.c +++ b/numpy/core/src/multiarray/refcount.c @@ -11,6 +11,7 @@ #define _MULTIARRAYMODULE #include "numpy/arrayobject.h" #include "numpy/arrayscalars.h" +#include "iterators.h" #include "npy_config.h" @@ -210,21 +211,22 @@ PyArray_XDECREF(PyArrayObject *mp) npy_intp i, n; PyObject **data; PyObject *temp; - PyArrayIterObject *it; + /* + * statically allocating it allows this function to not modify the + * reference count of the array for use during dealloc. + * (statically is not necessary as such) + */ + PyArrayIterObject it; if (!PyDataType_REFCHK(PyArray_DESCR(mp))) { return 0; } if (PyArray_DESCR(mp)->type_num != NPY_OBJECT) { - it = (PyArrayIterObject *)PyArray_IterNew((PyObject *)mp); - if (it == NULL) { - return -1; + PyArray_RawIterBaseInit(&it, mp); + while(it.index < it.size) { + PyArray_Item_XDECREF(it.dataptr, PyArray_DESCR(mp)); + PyArray_ITER_NEXT(&it); } - while(it->index < it->size) { - PyArray_Item_XDECREF(it->dataptr, PyArray_DESCR(mp)); - PyArray_ITER_NEXT(it); - } - Py_DECREF(it); return 0; } @@ -242,16 +244,12 @@ PyArray_XDECREF(PyArrayObject *mp) } } else { /* handles misaligned data too */ - it = (PyArrayIterObject *)PyArray_IterNew((PyObject *)mp); - if (it == NULL) { - return -1; - } - while(it->index < it->size) { - NPY_COPY_PYOBJECT_PTR(&temp, it->dataptr); + PyArray_RawIterBaseInit(&it, mp); + while(it.index < it.size) { + NPY_COPY_PYOBJECT_PTR(&temp, it.dataptr); Py_XDECREF(temp); - PyArray_ITER_NEXT(it); + PyArray_ITER_NEXT(&it); } - Py_DECREF(it); } return 0; } diff --git a/numpy/core/src/multiarray/scalartypes.c.src b/numpy/core/src/multiarray/scalartypes.c.src index cfb21f50e..5da7f7738 100644 --- a/numpy/core/src/multiarray/scalartypes.c.src +++ b/numpy/core/src/multiarray/scalartypes.c.src @@ -4060,8 +4060,11 @@ initialize_casting_tables(void) _npy_can_cast_safely_table[_FROM_NUM][NPY_STRING] = 1; _npy_can_cast_safely_table[_FROM_NUM][NPY_UNICODE] = 1; - /* Allow casts from any integer to the TIMEDELTA type */ -#if @from_isint@ || @from_isuint@ +#if @from_isint@ && NPY_SIZEOF_TIMEDELTA >= _FROM_BSIZE + /* Allow casts from smaller or equal signed integers to the TIMEDELTA type */ + _npy_can_cast_safely_table[_FROM_NUM][NPY_TIMEDELTA] = 1; +#elif @from_isuint@ && NPY_SIZEOF_TIMEDELTA > _FROM_BSIZE + /* Allow casts from smaller unsigned integers to the TIMEDELTA type */ _npy_can_cast_safely_table[_FROM_NUM][NPY_TIMEDELTA] = 1; #endif @@ -4492,6 +4495,36 @@ initialize_numeric_types(void) PyArrayIter_Type.tp_iter = PyObject_SelfIter; PyArrayMapIter_Type.tp_iter = PyObject_SelfIter; + + /* + * Give types different names when they are the same size (gh-9799). + * `np.intX` always refers to the first int of that size in the sequence + * `['LONG', 'LONGLONG', 'INT', 'SHORT', 'BYTE']`. + */ +#if (NPY_SIZEOF_BYTE == NPY_SIZEOF_SHORT) + PyByteArrType_Type.tp_name = "numpy.byte"; + PyUByteArrType_Type.tp_name = "numpy.ubyte"; +#endif +#if (NPY_SIZEOF_SHORT == NPY_SIZEOF_INT) + PyShortArrType_Type.tp_name = "numpy.short"; + PyUShortArrType_Type.tp_name = "numpy.ushort"; +#endif +#if (NPY_SIZEOF_INT == NPY_SIZEOF_LONG) + PyIntArrType_Type.tp_name = "numpy.intc"; + PyUIntArrType_Type.tp_name = "numpy.uintc"; +#endif +#if (NPY_SIZEOF_LONGLONG == NPY_SIZEOF_LONG) + PyLongLongArrType_Type.tp_name = "numpy.longlong"; + PyULongLongArrType_Type.tp_name = "numpy.ulonglong"; +#endif + + /* + Do the same for longdouble + */ +#if (NPY_SIZEOF_LONGDOUBLE == NPY_SIZEOF_DOUBLE) + PyLongDoubleArrType_Type.tp_name = "numpy.longdouble"; + PyCLongDoubleArrType_Type.tp_name = "numpy.clongdouble"; +#endif } typedef struct { diff --git a/numpy/core/src/multiarray/shape.c b/numpy/core/src/multiarray/shape.c index 1dffdeaed..4e31f003b 100644 --- a/numpy/core/src/multiarray/shape.c +++ b/numpy/core/src/multiarray/shape.c @@ -26,7 +26,7 @@ static int _fix_unknown_dimension(PyArray_Dims *newshape, PyArrayObject *arr); static int -_attempt_nocopy_reshape(PyArrayObject *self, int newnd, npy_intp* newdims, +_attempt_nocopy_reshape(PyArrayObject *self, int newnd, const npy_intp *newdims, npy_intp *newstrides, int is_f_order); static void @@ -40,7 +40,7 @@ _putzero(char *optr, PyObject *zero, PyArray_Descr *dtype); */ NPY_NO_EXPORT PyObject * PyArray_Resize(PyArrayObject *self, PyArray_Dims *newshape, int refcheck, - NPY_ORDER order) + NPY_ORDER NPY_UNUSED(order)) { npy_intp oldnbytes, newnbytes; npy_intp oldsize, newsize; @@ -361,7 +361,7 @@ _putzero(char *optr, PyObject *zero, PyArray_Descr *dtype) * stride of the next-fastest index. */ static int -_attempt_nocopy_reshape(PyArrayObject *self, int newnd, npy_intp* newdims, +_attempt_nocopy_reshape(PyArrayObject *self, int newnd, const npy_intp *newdims, npy_intp *newstrides, int is_f_order) { int oldnd; @@ -766,7 +766,7 @@ static int _npy_stride_sort_item_comparator(const void *a, const void *b) * [(2, 12), (0, 4), (1, -2)]. */ NPY_NO_EXPORT void -PyArray_CreateSortedStridePerm(int ndim, npy_intp *strides, +PyArray_CreateSortedStridePerm(int ndim, npy_intp const *strides, npy_stride_sort_item *out_strideperm) { int i; @@ -1048,7 +1048,7 @@ build_shape_string(npy_intp n, npy_intp *vals) * from a reduction result once its computation is complete. */ NPY_NO_EXPORT void -PyArray_RemoveAxesInPlace(PyArrayObject *arr, npy_bool *flags) +PyArray_RemoveAxesInPlace(PyArrayObject *arr, const npy_bool *flags) { PyArrayObject_fields *fa = (PyArrayObject_fields *)arr; npy_intp *shape = fa->dimensions, *strides = fa->strides; diff --git a/numpy/core/src/multiarray/vdot.c b/numpy/core/src/multiarray/vdot.c index 424a21710..9b5d19522 100644 --- a/numpy/core/src/multiarray/vdot.c +++ b/numpy/core/src/multiarray/vdot.c @@ -15,17 +15,17 @@ CFLOAT_vdot(char *ip1, npy_intp is1, char *ip2, npy_intp is2, char *op, npy_intp n, void *NPY_UNUSED(ignore)) { #if defined(HAVE_CBLAS) - int is1b = blas_stride(is1, sizeof(npy_cfloat)); - int is2b = blas_stride(is2, sizeof(npy_cfloat)); + CBLAS_INT is1b = blas_stride(is1, sizeof(npy_cfloat)); + CBLAS_INT is2b = blas_stride(is2, sizeof(npy_cfloat)); if (is1b && is2b) { double sum[2] = {0., 0.}; /* double for stability */ while (n > 0) { - int chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; + CBLAS_INT chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; float tmp[2]; - cblas_cdotc_sub((int)n, ip1, is1b, ip2, is2b, tmp); + CBLAS_FUNC(cblas_cdotc_sub)((CBLAS_INT)n, ip1, is1b, ip2, is2b, tmp); sum[0] += (double)tmp[0]; sum[1] += (double)tmp[1]; /* use char strides here */ @@ -66,17 +66,17 @@ CDOUBLE_vdot(char *ip1, npy_intp is1, char *ip2, npy_intp is2, char *op, npy_intp n, void *NPY_UNUSED(ignore)) { #if defined(HAVE_CBLAS) - int is1b = blas_stride(is1, sizeof(npy_cdouble)); - int is2b = blas_stride(is2, sizeof(npy_cdouble)); + CBLAS_INT is1b = blas_stride(is1, sizeof(npy_cdouble)); + CBLAS_INT is2b = blas_stride(is2, sizeof(npy_cdouble)); if (is1b && is2b) { double sum[2] = {0., 0.}; /* double for stability */ while (n > 0) { - int chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; + CBLAS_INT chunk = n < NPY_CBLAS_CHUNK ? n : NPY_CBLAS_CHUNK; double tmp[2]; - cblas_zdotc_sub((int)n, ip1, is1b, ip2, is2b, tmp); + CBLAS_FUNC(cblas_zdotc_sub)((CBLAS_INT)n, ip1, is1b, ip2, is2b, tmp); sum[0] += (double)tmp[0]; sum[1] += (double)tmp[1]; /* use char strides here */ diff --git a/numpy/core/src/npymath/npy_math_complex.c.src b/numpy/core/src/npymath/npy_math_complex.c.src index dad381232..8c432e483 100644 --- a/numpy/core/src/npymath/npy_math_complex.c.src +++ b/numpy/core/src/npymath/npy_math_complex.c.src @@ -40,13 +40,14 @@ * flag in an efficient way. The flag is IEEE specific. See * https://github.com/freebsd/freebsd/blob/4c6378299/lib/msun/src/catrig.c#L42 */ +#if !defined(HAVE_CACOSF) || !defined(HAVE_CACOSL) || !defined(HAVE_CASINHF) || !defined(HAVE_CASINHL) #define raise_inexact() do { \ volatile npy_float NPY_UNUSED(junk) = 1 + tiny; \ } while (0) static const volatile npy_float tiny = 3.9443045e-31f; - +#endif /**begin repeat * #type = npy_float, npy_double, npy_longdouble# @@ -64,9 +65,6 @@ static const volatile npy_float tiny = 3.9443045e-31f; * Constants *=========================================================*/ static const @ctype@ c_1@c@ = {1.0@C@, 0.0}; -static const @ctype@ c_half@c@ = {0.5@C@, 0.0}; -static const @ctype@ c_i@c@ = {0.0, 1.0@C@}; -static const @ctype@ c_ihalf@c@ = {0.0, 0.5@C@}; /*========================================================== * Helper functions @@ -76,22 +74,6 @@ static const @ctype@ c_ihalf@c@ = {0.0, 0.5@C@}; *=========================================================*/ static NPY_INLINE @ctype@ -cadd@c@(@ctype@ a, @ctype@ b) -{ - return npy_cpack@c@(npy_creal@c@(a) + npy_creal@c@(b), - npy_cimag@c@(a) + npy_cimag@c@(b)); -} - -static NPY_INLINE -@ctype@ -csub@c@(@ctype@ a, @ctype@ b) -{ - return npy_cpack@c@(npy_creal@c@(a) - npy_creal@c@(b), - npy_cimag@c@(a) - npy_cimag@c@(b)); -} - -static NPY_INLINE -@ctype@ cmul@c@(@ctype@ a, @ctype@ b) { @type@ ar, ai, br, bi; @@ -132,20 +114,6 @@ cdiv@c@(@ctype@ a, @ctype@ b) } } -static NPY_INLINE -@ctype@ -cneg@c@(@ctype@ a) -{ - return npy_cpack@c@(-npy_creal@c@(a), -npy_cimag@c@(a)); -} - -static NPY_INLINE -@ctype@ -cmuli@c@(@ctype@ a) -{ - return npy_cpack@c@(-npy_cimag@c@(a), npy_creal@c@(a)); -} - /*========================================================== * Custom implementation of missing complex C99 functions *=========================================================*/ diff --git a/numpy/core/src/npymath/npy_math_internal.h.src b/numpy/core/src/npymath/npy_math_internal.h.src index fa820baac..18b6d1434 100644 --- a/numpy/core/src/npymath/npy_math_internal.h.src +++ b/numpy/core/src/npymath/npy_math_internal.h.src @@ -716,3 +716,44 @@ npy_@func@@c@(@type@ a, @type@ b) return npy_@func@u@c@(a < 0 ? -a : a, b < 0 ? -b : b); } /**end repeat**/ + +/* Unlike LCM and GCD, we need byte and short variants for the shift operators, + * since the result is dependent on the width of the type + */ +/**begin repeat + * + * #type = byte, short, int, long, longlong# + * #c = hh,h,,l,ll# + */ +/**begin repeat1 + * + * #u = u,# + * #is_signed = 0,1# + */ +NPY_INPLACE npy_@u@@type@ +npy_lshift@u@@c@(npy_@u@@type@ a, npy_@u@@type@ b) +{ + if (NPY_LIKELY((size_t)b < sizeof(a) * CHAR_BIT)) { + return a << b; + } + else { + return 0; + } +} +NPY_INPLACE npy_@u@@type@ +npy_rshift@u@@c@(npy_@u@@type@ a, npy_@u@@type@ b) +{ + if (NPY_LIKELY((size_t)b < sizeof(a) * CHAR_BIT)) { + return a >> b; + } +#if @is_signed@ + else if (a < 0) { + return (npy_@u@@type@)-1; /* preserve the sign bit */ + } +#endif + else { + return 0; + } +} +/**end repeat1**/ +/**end repeat**/ diff --git a/numpy/core/src/npysort/npysort_common.h b/numpy/core/src/npysort/npysort_common.h index 5fd03b96f..2a6e4d421 100644 --- a/numpy/core/src/npysort/npysort_common.h +++ b/numpy/core/src/npysort/npysort_common.h @@ -329,6 +329,14 @@ UNICODE_LT(const npy_ucs4 *s1, const npy_ucs4 *s2, size_t len) NPY_INLINE static int DATETIME_LT(npy_datetime a, npy_datetime b) { + if (a == NPY_DATETIME_NAT) { + return 0; + } + + if (b == NPY_DATETIME_NAT) { + return 1; + } + return a < b; } @@ -336,6 +344,14 @@ DATETIME_LT(npy_datetime a, npy_datetime b) NPY_INLINE static int TIMEDELTA_LT(npy_timedelta a, npy_timedelta b) { + if (a == NPY_DATETIME_NAT) { + return 0; + } + + if (b == NPY_DATETIME_NAT) { + return 1; + } + return a < b; } diff --git a/numpy/core/src/umath/_rational_tests.c.src b/numpy/core/src/umath/_rational_tests.c.src index 9e74845df..615e395c7 100644 --- a/numpy/core/src/umath/_rational_tests.c.src +++ b/numpy/core/src/umath/_rational_tests.c.src @@ -539,11 +539,11 @@ static PyObject* pyrational_str(PyObject* self) { rational x = ((PyRational*)self)->r; if (d(x)!=1) { - return PyString_FromFormat( + return PyUString_FromFormat( "%ld/%ld",(long)x.n,(long)d(x)); } else { - return PyString_FromFormat( + return PyUString_FromFormat( "%ld",(long)x.n); } } diff --git a/numpy/core/src/umath/cpuid.c b/numpy/core/src/umath/cpuid.c index 8673f1736..72c6493e8 100644 --- a/numpy/core/src/umath/cpuid.c +++ b/numpy/core/src/umath/cpuid.c @@ -48,6 +48,25 @@ int os_avx512_support(void) #endif } +static NPY_INLINE +int cpu_supports_fma(void) +{ +#ifdef __x86_64__ + unsigned int feature = 0x01; + unsigned int a, b, c, d; + __asm__ volatile ( + "cpuid" "\n\t" + : "=a" (a), "=b" (b), "=c" (c), "=d" (d) + : "a" (feature)); + /* + * FMA is the 12th bit of ECX + */ + return (c >> 12) & 1; +#else + return 0; +#endif +} + /* * Primitive cpu feature detect function * Currently only supports checking for avx on gcc compatible compilers. @@ -63,6 +82,9 @@ npy_cpu_supports(const char * feature) return 0; #endif } + else if (strcmp(feature, "fma") == 0) { + return cpu_supports_fma() && __builtin_cpu_supports("avx2") && os_avx_support(); + } else if (strcmp(feature, "avx2") == 0) { return __builtin_cpu_supports("avx2") && os_avx_support(); } diff --git a/numpy/core/src/umath/funcs.inc.src b/numpy/core/src/umath/funcs.inc.src index c2732f925..10ed66e50 100644 --- a/numpy/core/src/umath/funcs.inc.src +++ b/numpy/core/src/umath/funcs.inc.src @@ -161,7 +161,7 @@ npy_ObjectLogicalNot(PyObject *i1) static PyObject * npy_ObjectFloor(PyObject *obj) { - PyObject *math_floor_func = NULL; + static PyObject *math_floor_func = NULL; npy_cache_import("math", "floor", &math_floor_func); if (math_floor_func == NULL) { @@ -172,7 +172,7 @@ npy_ObjectFloor(PyObject *obj) { static PyObject * npy_ObjectCeil(PyObject *obj) { - PyObject *math_ceil_func = NULL; + static PyObject *math_ceil_func = NULL; npy_cache_import("math", "ceil", &math_ceil_func); if (math_ceil_func == NULL) { @@ -183,7 +183,7 @@ npy_ObjectCeil(PyObject *obj) { static PyObject * npy_ObjectTrunc(PyObject *obj) { - PyObject *math_trunc_func = NULL; + static PyObject *math_trunc_func = NULL; npy_cache_import("math", "trunc", &math_trunc_func); if (math_trunc_func == NULL) { @@ -228,7 +228,8 @@ npy_ObjectGCD(PyObject *i1, PyObject *i2) return NULL; } /* _gcd has some unusual behaviour regarding sign */ - return PyNumber_Absolute(gcd); + Py_SETREF(gcd, PyNumber_Absolute(gcd)); + return gcd; } } @@ -246,17 +247,19 @@ npy_ObjectLCM(PyObject *i1, PyObject *i2) * no remainder */ tmp = PyNumber_FloorDivide(i1, gcd); + Py_DECREF(gcd); if(tmp == NULL) { return NULL; } - tmp = PyNumber_Multiply(tmp, i2); + Py_SETREF(tmp, PyNumber_Multiply(tmp, i2)); if(tmp == NULL) { return NULL; } /* even though we fix gcd to be positive, we need to do it again here */ - return PyNumber_Absolute(tmp); + Py_SETREF(tmp, PyNumber_Absolute(tmp)); + return tmp; } diff --git a/numpy/core/src/umath/loops.c.src b/numpy/core/src/umath/loops.c.src index 1931cd100..8a2e5bc40 100644 --- a/numpy/core/src/umath/loops.c.src +++ b/numpy/core/src/umath/loops.c.src @@ -54,210 +54,123 @@ ** GENERIC FLOAT LOOPS ** *****************************************************************************/ +/* direct loops using a suitable callback */ -typedef float halfUnaryFunc(npy_half x); -typedef float floatUnaryFunc(float x); -typedef double doubleUnaryFunc(double x); -typedef npy_longdouble longdoubleUnaryFunc(npy_longdouble x); -typedef npy_half halfBinaryFunc(npy_half x, npy_half y); -typedef float floatBinaryFunc(float x, float y); -typedef double doubleBinaryFunc(double x, double y); -typedef npy_longdouble longdoubleBinaryFunc(npy_longdouble x, npy_longdouble y); - +/**begin repeat + * #c = e, f, d, g# + * #type = npy_half, npy_float, npy_double, npy_longdouble# + **/ /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_e_e(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c@_@c@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - halfUnaryFunc *f = (halfUnaryFunc *)func; + typedef @type@ func_type(@type@); + func_type *f = (func_type *)func; UNARY_LOOP { - const npy_half in1 = *(npy_half *)ip1; - *(npy_half *)op1 = f(in1); + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = f(in1); } } /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_e_e_As_f_f(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c@@c@_@c@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - floatUnaryFunc *f = (floatUnaryFunc *)func; - UNARY_LOOP { - const float in1 = npy_half_to_float(*(npy_half *)ip1); - *(npy_half *)op1 = npy_float_to_half(f(in1)); + typedef @type@ func_type(@type@, @type@); + func_type *f = (func_type *)func; + BINARY_LOOP { + @type@ in1 = *(@type@ *)ip1; + @type@ in2 = *(@type@ *)ip2; + *(@type@ *)op1 = f(in1, in2); } } -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_e_e_As_d_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - doubleUnaryFunc *f = (doubleUnaryFunc *)func; - UNARY_LOOP { - const double in1 = npy_half_to_double(*(npy_half *)ip1); - *(npy_half *)op1 = npy_double_to_half(f(in1)); - } -} +/**end repeat**/ -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_f_f(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - floatUnaryFunc *f = (floatUnaryFunc *)func; - UNARY_LOOP { - const float in1 = *(float *)ip1; - *(float *)op1 = f(in1); - } -} +/* indirect loops with casting */ +/**begin repeat + * #c1 = e, e, f# + * #type1 = npy_half, npy_half, npy_float# + * #c2 = f, d, d# + * #type2 = npy_float, npy_double, npy_double# + * + * #conv12 = npy_half_to_float, npy_half_to_double, (double)# + * #conv21 = npy_float_to_half, npy_double_to_half, (float)# + **/ /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_f_f_As_d_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c1@_@c1@_As_@c2@_@c2@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - doubleUnaryFunc *f = (doubleUnaryFunc *)func; + typedef @type2@ func_type(@type2@); + func_type *f = (func_type *)func; UNARY_LOOP { - const float in1 = *(float *)ip1; - *(float *)op1 = (float)f((double)in1); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_ee_e(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - halfBinaryFunc *f = (halfBinaryFunc *)func; - BINARY_LOOP { - npy_half in1 = *(npy_half *)ip1; - npy_half in2 = *(npy_half *)ip2; - *(npy_half *)op1 = f(in1, in2); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_ee_e_As_ff_f(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - floatBinaryFunc *f = (floatBinaryFunc *)func; - BINARY_LOOP { - float in1 = npy_half_to_float(*(npy_half *)ip1); - float in2 = npy_half_to_float(*(npy_half *)ip2); - *(npy_half *)op1 = npy_float_to_half(f(in1, in2)); + const @type2@ in1 = @conv12@(*(@type1@ *)ip1); + *(@type1@ *)op1 = @conv21@(f(in1)); } } - /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_ee_e_As_dd_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c1@@c1@_@c1@_As_@c2@@c2@_@c2@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - doubleBinaryFunc *f = (doubleBinaryFunc *)func; + typedef @type2@ func_type(@type2@, @type2@); + func_type *f = (func_type *)func; BINARY_LOOP { - double in1 = npy_half_to_double(*(npy_half *)ip1); - double in2 = npy_half_to_double(*(npy_half *)ip2); - *(npy_half *)op1 = npy_double_to_half(f(in1, in2)); + const @type2@ in1 = @conv12@(*(@type1@ *)ip1); + const @type2@ in2 = @conv12@(*(@type1@ *)ip2); + *(@type1@ *)op1 = @conv21@(f(in1, in2)); } } -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_ff_f(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - floatBinaryFunc *f = (floatBinaryFunc *)func; - BINARY_LOOP { - float in1 = *(float *)ip1; - float in2 = *(float *)ip2; - *(float *)op1 = f(in1, in2); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_ff_f_As_dd_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - doubleBinaryFunc *f = (doubleBinaryFunc *)func; - BINARY_LOOP { - float in1 = *(float *)ip1; - float in2 = *(float *)ip2; - *(float *)op1 = (double)f((double)in1, (double)in2); - } -} +/**end repeat**/ -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_d_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - doubleUnaryFunc *f = (doubleUnaryFunc *)func; - UNARY_LOOP { - double in1 = *(double *)ip1; - *(double *)op1 = f(in1); - } -} +/****************************************************************************** + ** GENERIC COMPLEX LOOPS ** + *****************************************************************************/ -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_dd_d(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - doubleBinaryFunc *f = (doubleBinaryFunc *)func; - BINARY_LOOP { - double in1 = *(double *)ip1; - double in2 = *(double *)ip2; - *(double *)op1 = f(in1, in2); - } -} +/* direct loops using a suitable callback */ +/**begin repeat + * #c = F, D, G# + * #type = npy_cfloat, npy_cdouble, npy_clongdouble# + **/ /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_g_g(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c@_@c@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - longdoubleUnaryFunc *f = (longdoubleUnaryFunc *)func; + typedef void func_type(@type@ *, @type@ *); + func_type *f = (func_type *)func; UNARY_LOOP { - npy_longdouble in1 = *(npy_longdouble *)ip1; - *(npy_longdouble *)op1 = f(in1); + @type@ in1 = *(@type@ *)ip1; + @type@ *out = (@type@ *)op1; + f(&in1, out); } } /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_gg_g(char **args, npy_intp *dimensions, npy_intp *steps, void *func) +PyUFunc_@c@@c@_@c@(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - longdoubleBinaryFunc *f = (longdoubleBinaryFunc *)func; + typedef void func_type(@type@ *, @type@ *, @type@ *); + func_type *f = (func_type *)func; BINARY_LOOP { - npy_longdouble in1 = *(npy_longdouble *)ip1; - npy_longdouble in2 = *(npy_longdouble *)ip2; - *(npy_longdouble *)op1 = f(in1, in2); + @type@ in1 = *(@type@ *)ip1; + @type@ in2 = *(@type@ *)ip2; + @type@ *out = (@type@ *)op1; + f(&in1, &in2, out); } } +/**end repeat**/ - -/****************************************************************************** - ** GENERIC COMPLEX LOOPS ** - *****************************************************************************/ - - -typedef void cdoubleUnaryFunc(npy_cdouble *x, npy_cdouble *r); -typedef void cfloatUnaryFunc(npy_cfloat *x, npy_cfloat *r); -typedef void clongdoubleUnaryFunc(npy_clongdouble *x, npy_clongdouble *r); -typedef void cdoubleBinaryFunc(npy_cdouble *x, npy_cdouble *y, npy_cdouble *r); -typedef void cfloatBinaryFunc(npy_cfloat *x, npy_cfloat *y, npy_cfloat *r); -typedef void clongdoubleBinaryFunc(npy_clongdouble *x, npy_clongdouble *y, - npy_clongdouble *r); - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_F_F(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - cfloatUnaryFunc *f = (cfloatUnaryFunc *)func; - UNARY_LOOP { - npy_cfloat in1 = *(npy_cfloat *)ip1; - npy_cfloat *out = (npy_cfloat *)op1; - f(&in1, out); - } -} - +/* indirect loops with casting */ /*UFUNC_API*/ NPY_NO_EXPORT void PyUFunc_F_F_As_D_D(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - cdoubleUnaryFunc *f = (cdoubleUnaryFunc *)func; + typedef void func_type(npy_cdouble *, npy_cdouble *); + func_type *f = (func_type *)func; UNARY_LOOP { npy_cdouble tmp, out; tmp.real = (double)((float *)ip1)[0]; @@ -270,22 +183,10 @@ PyUFunc_F_F_As_D_D(char **args, npy_intp *dimensions, npy_intp *steps, void *fun /*UFUNC_API*/ NPY_NO_EXPORT void -PyUFunc_FF_F(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - cfloatBinaryFunc *f = (cfloatBinaryFunc *)func; - BINARY_LOOP { - npy_cfloat in1 = *(npy_cfloat *)ip1; - npy_cfloat in2 = *(npy_cfloat *)ip2; - npy_cfloat *out = (npy_cfloat *)op1; - f(&in1, &in2, out); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void PyUFunc_FF_F_As_DD_D(char **args, npy_intp *dimensions, npy_intp *steps, void *func) { - cdoubleBinaryFunc *f = (cdoubleBinaryFunc *)func; + typedef void func_type(npy_cdouble *, npy_cdouble *, npy_cdouble *); + func_type *f = (func_type *)func; BINARY_LOOP { npy_cdouble tmp1, tmp2, out; tmp1.real = (double)((float *)ip1)[0]; @@ -298,56 +199,6 @@ PyUFunc_FF_F_As_DD_D(char **args, npy_intp *dimensions, npy_intp *steps, void *f } } -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_D_D(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - cdoubleUnaryFunc *f = (cdoubleUnaryFunc *)func; - UNARY_LOOP { - npy_cdouble in1 = *(npy_cdouble *)ip1; - npy_cdouble *out = (npy_cdouble *)op1; - f(&in1, out); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_DD_D(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - cdoubleBinaryFunc *f = (cdoubleBinaryFunc *)func; - BINARY_LOOP { - npy_cdouble in1 = *(npy_cdouble *)ip1; - npy_cdouble in2 = *(npy_cdouble *)ip2; - npy_cdouble *out = (npy_cdouble *)op1; - f(&in1, &in2, out); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_G_G(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - clongdoubleUnaryFunc *f = (clongdoubleUnaryFunc *)func; - UNARY_LOOP { - npy_clongdouble in1 = *(npy_clongdouble *)ip1; - npy_clongdouble *out = (npy_clongdouble *)op1; - f(&in1, out); - } -} - -/*UFUNC_API*/ -NPY_NO_EXPORT void -PyUFunc_GG_G(char **args, npy_intp *dimensions, npy_intp *steps, void *func) -{ - clongdoubleBinaryFunc *f = (clongdoubleBinaryFunc *)func; - BINARY_LOOP { - npy_clongdouble in1 = *(npy_clongdouble *)ip1; - npy_clongdouble in2 = *(npy_clongdouble *)ip2; - npy_clongdouble *out = (npy_clongdouble *)op1; - f(&in1, &in2, out); - } -} - /****************************************************************************** ** GENERIC OBJECT lOOPS ** @@ -398,6 +249,7 @@ PyUFunc_O_O_method(char **args, npy_intp *dimensions, npy_intp *steps, void *fun i, type->tp_name, meth); npy_PyErr_ChainExceptionsCause(exc, val, tb); Py_DECREF(tup); + Py_XDECREF(func); return; } ret = PyObject_Call(func, tup, NULL); @@ -702,6 +554,7 @@ BOOL_@kind@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED * #ftype = npy_float, npy_float, npy_float, npy_float, npy_double, npy_double, * npy_double, npy_double, npy_double, npy_double# * #SIGNED = 1, 0, 1, 0, 1, 0, 1, 0, 1, 0# + * #c = hh,uhh,h,uh,,u,l,ul,ll,ull# */ #define @TYPE@_floor_divide @TYPE@_divide @@ -779,16 +632,15 @@ NPY_NO_EXPORT NPY_GCC_OPT_3 @ATTR@ void /**begin repeat2 * Arithmetic - * #kind = add, subtract, multiply, bitwise_and, bitwise_or, bitwise_xor, - * left_shift, right_shift# - * #OP = +, -,*, &, |, ^, <<, >># + * #kind = add, subtract, multiply, bitwise_and, bitwise_or, bitwise_xor# + * #OP = +, -, *, &, |, ^# */ #if @CHK@ NPY_NO_EXPORT NPY_GCC_OPT_3 @ATTR@ void @TYPE@_@kind@@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)) { - if(IS_BINARY_REDUCE) { + if (IS_BINARY_REDUCE) { BINARY_REDUCE_LOOP(@type@) { io1 @OP@= *(@type@ *)ip2; } @@ -802,6 +654,47 @@ NPY_NO_EXPORT NPY_GCC_OPT_3 @ATTR@ void /**end repeat2**/ +/* + * Arithmetic bit shift operations. + * + * Intel hardware masks bit shift values, so large shifts wrap around + * and can produce surprising results. The special handling ensures that + * behavior is independent of compiler or hardware. + * TODO: We could implement consistent behavior for negative shifts, + * which is undefined in C. + */ + +#define INT_left_shift_needs_clear_floatstatus +#define UINT_left_shift_needs_clear_floatstatus + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_left_shift@isa@(char **args, npy_intp *dimensions, npy_intp *steps, + void *NPY_UNUSED(func)) +{ + BINARY_LOOP_FAST(@type@, @type@, *out = npy_lshift@c@(in1, in2)); + +#ifdef @TYPE@_left_shift_needs_clear_floatstatus + // For some reason, our macOS CI sets an "invalid" flag here, but only + // for some types. + npy_clear_floatstatus_barrier((char*)dimensions); +#endif +} + +#undef INT_left_shift_needs_clear_floatstatus +#undef UINT_left_shift_needs_clear_floatstatus + +NPY_NO_EXPORT +#ifndef NPY_DO_NOT_OPTIMIZE_@TYPE@_right_shift +NPY_GCC_OPT_3 +#endif +void +@TYPE@_right_shift@isa@(char **args, npy_intp *dimensions, npy_intp *steps, + void *NPY_UNUSED(func)) +{ + BINARY_LOOP_FAST(@type@, @type@, *out = npy_rshift@c@(in1, in2)); +} + + /**begin repeat2 * #kind = equal, not_equal, greater, greater_equal, less, less_equal, * logical_and, logical_or# @@ -1208,6 +1101,12 @@ NPY_NO_EXPORT void } NPY_NO_EXPORT void +@TYPE@_isinf(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)) +{ + UNARY_LOOP_FAST(npy_bool, npy_bool, (void)in; *out = NPY_FALSE); +} + +NPY_NO_EXPORT void @TYPE@__ones_like(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) { OUTPUT_LOOP { @@ -1256,13 +1155,36 @@ NPY_NO_EXPORT void const @type@ in1 = *(@type@ *)ip1; const @type@ in2 = *(@type@ *)ip2; if (in1 == NPY_DATETIME_NAT) { + *((@type@ *)op1) = in1; + } + else if (in2 == NPY_DATETIME_NAT) { + *((@type@ *)op1) = in2; + } + else { + *((@type@ *)op1) = (in1 @OP@ in2) ? in1 : in2; + } + } +} +/**end repeat1**/ + +/**begin repeat1 + * #kind = fmax, fmin# + * #OP = >=, <=# + **/ +NPY_NO_EXPORT void +@TYPE@_@kind@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)) +{ + BINARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + const @type@ in2 = *(@type@ *)ip2; + if (in1 == NPY_DATETIME_NAT) { *((@type@ *)op1) = in2; } else if (in2 == NPY_DATETIME_NAT) { *((@type@ *)op1) = in1; } else { - *((@type@ *)op1) = (in1 @OP@ in2) ? in1 : in2; + *((@type@ *)op1) = in1 @OP@ in2 ? in1 : in2; } } } @@ -1597,8 +1519,32 @@ NPY_NO_EXPORT void /**end repeat**/ /**begin repeat - * #func = exp, log# - * #scalarf = npy_expf, npy_logf# + * #func = rint, ceil, floor, trunc# + * #scalarf = npy_rint, npy_ceil, npy_floor, npy_trunc# + */ + +/**begin repeat1 +* #TYPE = FLOAT, DOUBLE# +* #type = npy_float, npy_double# +* #typesub = f, # +*/ + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_@func@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = @scalarf@@typesub@(in1); + } +} + + +/**end repeat1**/ +/**end repeat**/ + +/**begin repeat + * #func = sin, cos, exp, log# + * #scalarf = npy_sinf, npy_cosf, npy_expf, npy_logf# */ NPY_NO_EXPORT NPY_GCC_OPT_3 void @@ -1613,12 +1559,84 @@ FLOAT_@func@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSE /**end repeat**/ /**begin repeat - * #isa = avx512f, avx2# - * #ISA = AVX512F, AVX2# + * #isa = avx512f, fma# + * #ISA = AVX512F, FMA# * #CHK = HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS# */ /**begin repeat1 + * #TYPE = FLOAT, DOUBLE# + * #type = npy_float, npy_double# + * #typesub = f, # + */ + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_sqrt_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_sqrt_@TYPE@(args, dimensions, steps)) { + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = npy_sqrt@typesub@(in1); + } + } +} + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_absolute_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_absolute_@TYPE@(args, dimensions, steps)) { + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + const @type@ tmp = in1 > 0 ? in1 : -in1; + /* add 0 to clear -0.0 */ + *((@type@ *)op1) = tmp + 0; + } + } + npy_clear_floatstatus_barrier((char*)dimensions); +} + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_square_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_square_@TYPE@(args, dimensions, steps)) { + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = in1*in1; + } + } +} + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_reciprocal_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_reciprocal_@TYPE@(args, dimensions, steps)) { + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = 1.0f/in1; + } + } +} + +/**begin repeat2 + * #func = rint, ceil, floor, trunc# + * #scalarf = npy_rint, npy_ceil, npy_floor, npy_trunc# + */ + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_@func@_@TYPE@(args, dimensions, steps)) { + UNARY_LOOP { + const @type@ in1 = *(@type@ *)ip1; + *(@type@ *)op1 = @scalarf@@typesub@(in1); + } + } +} + +/**end repeat2**/ +/**end repeat1**/ + +/**begin repeat1 * #func = exp, log# * #scalarf = npy_expf, npy_logf# */ @@ -1645,8 +1663,32 @@ FLOAT_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY } /**end repeat1**/ + +/**begin repeat1 + * #func = cos, sin# + * #enum = npy_compute_cos, npy_compute_sin# + * #scalarf = npy_cosf, npy_sinf# + */ + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +FLOAT_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)) +{ + if (!run_unary_@isa@_sincos_FLOAT(args, dimensions, steps, @enum@)) { + UNARY_LOOP { +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS + @ISA@_sincos_FLOAT((npy_float *)op1, (npy_float *)ip1, 1, steps[0], @enum@); +#else + const npy_float in1 = *(npy_float *)ip1; + *(npy_float *)op1 = @scalarf@(in1); +#endif + } + } +} + +/**end repeat1**/ /**end repeat**/ + /**begin repeat * Float types * #type = npy_float, npy_double, npy_longdouble, npy_float# diff --git a/numpy/core/src/umath/loops.h.src b/numpy/core/src/umath/loops.h.src index 7f05a693a..7558de0bb 100644 --- a/numpy/core/src/umath/loops.h.src +++ b/numpy/core/src/umath/loops.h.src @@ -7,14 +7,12 @@ #define _NPY_UMATH_LOOPS_H_ #define BOOL_invert BOOL_logical_not -#define BOOL_negative BOOL_logical_not #define BOOL_add BOOL_logical_or #define BOOL_bitwise_and BOOL_logical_and #define BOOL_bitwise_or BOOL_logical_or #define BOOL_logical_xor BOOL_not_equal #define BOOL_bitwise_xor BOOL_logical_xor #define BOOL_multiply BOOL_logical_and -#define BOOL_subtract BOOL_logical_xor #define BOOL_maximum BOOL_logical_or #define BOOL_minimum BOOL_logical_and #define BOOL_fmax BOOL_maximum @@ -175,16 +173,29 @@ NPY_NO_EXPORT void */ NPY_NO_EXPORT void @TYPE@_sqrt(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); + +/**begin repeat1 + * #isa = avx512f, fma# + */ + +/**begin repeat2 + * #func = sqrt, absolute, square, reciprocal# + */ +NPY_NO_EXPORT void +@TYPE@_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); + +/**end repeat2**/ +/**end repeat1**/ /**end repeat**/ /**begin repeat - * #func = exp, log# + * #func = sin, cos, exp, log# */ NPY_NO_EXPORT void FLOAT_@func@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); /**begin repeat1 - * #isa = avx512f, avx2# + * #isa = avx512f, fma# */ NPY_NO_EXPORT void @@ -194,6 +205,26 @@ FLOAT_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY /**end repeat**/ /**begin repeat + * #func = rint, ceil, floor, trunc# + */ + +/**begin repeat1 +* #TYPE = FLOAT, DOUBLE# +*/ + +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_@func@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)); + +/**begin repeat2 + * #isa = avx512f, fma# + */ +NPY_NO_EXPORT NPY_GCC_OPT_3 void +@TYPE@_@func@_@isa@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)); +/**end repeat2**/ +/**end repeat1**/ +/**end repeat**/ + +/**begin repeat * Float types * #TYPE = HALF, FLOAT, DOUBLE, LONGDOUBLE# * #c = f, f, , l# @@ -447,6 +478,11 @@ NPY_NO_EXPORT void @TYPE@_isfinite(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); NPY_NO_EXPORT void +@TYPE@_isinf(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); + +#define @TYPE@_isnan @TYPE@_isnat + +NPY_NO_EXPORT void @TYPE@__ones_like(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(data)); /**begin repeat1 @@ -458,8 +494,7 @@ NPY_NO_EXPORT void /**end repeat1**/ /**begin repeat1 - * #kind = maximum, minimum# - * #OP = >, <# + * #kind = maximum, minimum, fmin, fmax# **/ NPY_NO_EXPORT void @TYPE@_@kind@(char **args, npy_intp *dimensions, npy_intp *steps, void *NPY_UNUSED(func)); @@ -523,10 +558,6 @@ TIMEDELTA_mm_qm_divmod(char **args, npy_intp *dimensions, npy_intp *steps, void #define TIMEDELTA_mq_m_floor_divide TIMEDELTA_mq_m_divide #define TIMEDELTA_md_m_floor_divide TIMEDELTA_md_m_divide /* #define TIMEDELTA_mm_d_floor_divide TIMEDELTA_mm_d_divide */ -#define TIMEDELTA_fmin TIMEDELTA_minimum -#define TIMEDELTA_fmax TIMEDELTA_maximum -#define DATETIME_fmin DATETIME_minimum -#define DATETIME_fmax DATETIME_maximum /* ***************************************************************************** diff --git a/numpy/core/src/umath/matmul.c.src b/numpy/core/src/umath/matmul.c.src index b5204eca5..c8f7c654d 100644 --- a/numpy/core/src/umath/matmul.c.src +++ b/numpy/core/src/umath/matmul.c.src @@ -31,7 +31,11 @@ * -1 to be conservative, in case blas internally uses a for loop with an * inclusive upper bound */ +#ifndef HAVE_BLAS_ILP64 #define BLAS_MAXSIZE (NPY_MAX_INT - 1) +#else +#define BLAS_MAXSIZE (NPY_MAX_INT64 - 1) +#endif /* * Determine if a 2d matrix can be used by BLAS @@ -84,25 +88,25 @@ NPY_NO_EXPORT void * op: data in c order, m shape */ enum CBLAS_ORDER order; - int M, N, lda; + CBLAS_INT M, N, lda; assert(m <= BLAS_MAXSIZE && n <= BLAS_MAXSIZE); assert (is_blasable2d(is2_n, sizeof(@typ@), n, 1, sizeof(@typ@))); - M = (int)m; - N = (int)n; + M = (CBLAS_INT)m; + N = (CBLAS_INT)n; if (is_blasable2d(is1_m, is1_n, m, n, sizeof(@typ@))) { order = CblasColMajor; - lda = (int)(is1_m / sizeof(@typ@)); + lda = (CBLAS_INT)(is1_m / sizeof(@typ@)); } else { /* If not ColMajor, caller should have ensured we are RowMajor */ /* will not assert in release mode */ order = CblasRowMajor; assert(is_blasable2d(is1_n, is1_m, n, m, sizeof(@typ@))); - lda = (int)(is1_n / sizeof(@typ@)); + lda = (CBLAS_INT)(is1_n / sizeof(@typ@)); } - cblas_@prefix@gemv(order, CblasTrans, N, M, @step1@, ip1, lda, ip2, + CBLAS_FUNC(cblas_@prefix@gemv)(order, CblasTrans, N, M, @step1@, ip1, lda, ip2, is2_n / sizeof(@typ@), @step0@, op, op_m / sizeof(@typ@)); } @@ -117,37 +121,37 @@ NPY_NO_EXPORT void */ enum CBLAS_ORDER order = CblasRowMajor; enum CBLAS_TRANSPOSE trans1, trans2; - int M, N, P, lda, ldb, ldc; + CBLAS_INT M, N, P, lda, ldb, ldc; assert(m <= BLAS_MAXSIZE && n <= BLAS_MAXSIZE && p <= BLAS_MAXSIZE); - M = (int)m; - N = (int)n; - P = (int)p; + M = (CBLAS_INT)m; + N = (CBLAS_INT)n; + P = (CBLAS_INT)p; assert(is_blasable2d(os_m, os_p, m, p, sizeof(@typ@))); - ldc = (int)(os_m / sizeof(@typ@)); + ldc = (CBLAS_INT)(os_m / sizeof(@typ@)); if (is_blasable2d(is1_m, is1_n, m, n, sizeof(@typ@))) { trans1 = CblasNoTrans; - lda = (int)(is1_m / sizeof(@typ@)); + lda = (CBLAS_INT)(is1_m / sizeof(@typ@)); } else { /* If not ColMajor, caller should have ensured we are RowMajor */ /* will not assert in release mode */ assert(is_blasable2d(is1_n, is1_m, n, m, sizeof(@typ@))); trans1 = CblasTrans; - lda = (int)(is1_n / sizeof(@typ@)); + lda = (CBLAS_INT)(is1_n / sizeof(@typ@)); } if (is_blasable2d(is2_n, is2_p, n, p, sizeof(@typ@))) { trans2 = CblasNoTrans; - ldb = (int)(is2_n / sizeof(@typ@)); + ldb = (CBLAS_INT)(is2_n / sizeof(@typ@)); } else { /* If not ColMajor, caller should have ensured we are RowMajor */ /* will not assert in release mode */ assert(is_blasable2d(is2_p, is2_n, p, n, sizeof(@typ@))); trans2 = CblasTrans; - ldb = (int)(is2_p / sizeof(@typ@)); + ldb = (CBLAS_INT)(is2_p / sizeof(@typ@)); } /* * Use syrk if we have a case of a matrix times its transpose. @@ -162,12 +166,14 @@ NPY_NO_EXPORT void ) { npy_intp i,j; if (trans1 == CblasNoTrans) { - cblas_@prefix@syrk(order, CblasUpper, trans1, P, N, @step1@, - ip1, lda, @step0@, op, ldc); + CBLAS_FUNC(cblas_@prefix@syrk)( + order, CblasUpper, trans1, P, N, @step1@, + ip1, lda, @step0@, op, ldc); } else { - cblas_@prefix@syrk(order, CblasUpper, trans1, P, N, @step1@, - ip1, ldb, @step0@, op, ldc); + CBLAS_FUNC(cblas_@prefix@syrk)( + order, CblasUpper, trans1, P, N, @step1@, + ip1, ldb, @step0@, op, ldc); } /* Copy the triangle */ for (i = 0; i < P; i++) { @@ -178,8 +184,9 @@ NPY_NO_EXPORT void } else { - cblas_@prefix@gemm(order, trans1, trans2, M, P, N, @step1@, ip1, lda, - ip2, ldb, @step0@, op, ldc); + CBLAS_FUNC(cblas_@prefix@gemm)( + order, trans1, trans2, M, P, N, @step1@, ip1, lda, + ip2, ldb, @step0@, op, ldc); } } diff --git a/numpy/core/src/umath/override.c b/numpy/core/src/umath/override.c index 8d67f96ac..43bed425c 100644 --- a/numpy/core/src/umath/override.c +++ b/numpy/core/src/umath/override.c @@ -494,32 +494,18 @@ PyUFunc_CheckOverride(PyUFuncObject *ufunc, char *method, } else { /* not a tuple */ - if (nout > 1 && DEPRECATE("passing a single argument to the " - "'out' keyword argument of a " - "ufunc with\n" - "more than one output will " - "result in an error in the " - "future") < 0) { - /* - * If the deprecation is removed, also remove the loop - * below setting tuple items to None (but keep this future - * error message.) - */ + if (nout > 1) { PyErr_SetString(PyExc_TypeError, "'out' must be a tuple of arguments"); goto fail; } if (out != Py_None) { /* not already a tuple and not None */ - PyObject *out_tuple = PyTuple_New(nout); + PyObject *out_tuple = PyTuple_New(1); if (out_tuple == NULL) { goto fail; } - for (i = 1; i < nout; i++) { - Py_INCREF(Py_None); - PyTuple_SET_ITEM(out_tuple, i, Py_None); - } /* out was borrowed ref; make it permanent */ Py_INCREF(out); /* steals reference */ diff --git a/numpy/core/src/umath/reduction.c b/numpy/core/src/umath/reduction.c index fda2a12f6..4ce8d8ab7 100644 --- a/numpy/core/src/umath/reduction.c +++ b/numpy/core/src/umath/reduction.c @@ -36,7 +36,7 @@ * If 'dtype' isn't NULL, this function steals its reference. */ static PyArrayObject * -allocate_reduce_result(PyArrayObject *arr, npy_bool *axis_flags, +allocate_reduce_result(PyArrayObject *arr, const npy_bool *axis_flags, PyArray_Descr *dtype, int subok) { npy_intp strides[NPY_MAXDIMS], stride; @@ -84,7 +84,7 @@ allocate_reduce_result(PyArrayObject *arr, npy_bool *axis_flags, * The return value is a view into 'out'. */ static PyArrayObject * -conform_reduce_result(int ndim, npy_bool *axis_flags, +conform_reduce_result(int ndim, const npy_bool *axis_flags, PyArrayObject *out, int keepdims, const char *funcname, int need_copy) { @@ -251,7 +251,7 @@ PyArray_CreateReduceResult(PyArrayObject *operand, PyArrayObject *out, * Count the number of dimensions selected in 'axis_flags' */ static int -count_axes(int ndim, npy_bool *axis_flags) +count_axes(int ndim, const npy_bool *axis_flags) { int idim; int naxes = 0; @@ -299,7 +299,7 @@ count_axes(int ndim, npy_bool *axis_flags) NPY_NO_EXPORT PyArrayObject * PyArray_InitializeReduceResult( PyArrayObject *result, PyArrayObject *operand, - npy_bool *axis_flags, + const npy_bool *axis_flags, npy_intp *out_skip_first_count, const char *funcname) { npy_intp *strides, *shape, shape_orig[NPY_MAXDIMS]; @@ -528,7 +528,9 @@ PyUFunc_ReduceWrapper(PyArrayObject *operand, PyArrayObject *out, NPY_ITER_ALIGNED; if (wheremask != NULL) { op[2] = wheremask; - op_dtypes[2] = PyArray_DescrFromType(NPY_BOOL); + /* wheremask is guaranteed to be NPY_BOOL, so borrow its reference */ + op_dtypes[2] = PyArray_DESCR(wheremask); + assert(op_dtypes[2]->type_num == NPY_BOOL); if (op_dtypes[2] == NULL) { goto fail; } diff --git a/numpy/core/src/umath/scalarmath.c.src b/numpy/core/src/umath/scalarmath.c.src index a7987acda..d5d8d659b 100644 --- a/numpy/core/src/umath/scalarmath.c.src +++ b/numpy/core/src/umath/scalarmath.c.src @@ -246,25 +246,26 @@ static void /**end repeat**/ - -/* QUESTION: Should we check for overflow / underflow in (l,r)shift? */ - /**begin repeat * #name = byte, ubyte, short, ushort, int, uint, * long, ulong, longlong, ulonglong# * #type = npy_byte, npy_ubyte, npy_short, npy_ushort, npy_int, npy_uint, * npy_long, npy_ulong, npy_longlong, npy_ulonglong# + * #suffix = hh,uhh,h,uh,,u,l,ul,ll,ull# */ /**begin repeat1 - * #oper = and, xor, or, lshift, rshift# - * #op = &, ^, |, <<, >># + * #oper = and, xor, or# + * #op = &, ^, |# */ #define @name@_ctype_@oper@(arg1, arg2, out) *(out) = (arg1) @op@ (arg2) /**end repeat1**/ +#define @name@_ctype_lshift(arg1, arg2, out) *(out) = npy_lshift@suffix@(arg1, arg2) +#define @name@_ctype_rshift(arg1, arg2, out) *(out) = npy_rshift@suffix@(arg1, arg2) + /**end repeat**/ /**begin repeat @@ -405,21 +406,22 @@ half_ctype_divmod(npy_half a, npy_half b, npy_half *out1, npy_half *out2) { /**begin repeat * #name = float, double, longdouble# * #type = npy_float, npy_double, npy_longdouble# + * #c = f,,l# */ -static npy_@name@ (*_basic_@name@_pow)(@type@ a, @type@ b); static void @name@_ctype_power(@type@ a, @type@ b, @type@ *out) { - *out = _basic_@name@_pow(a, b); + *out = npy_pow@c@(a, b); } + /**end repeat**/ static void half_ctype_power(npy_half a, npy_half b, npy_half *out) { const npy_float af = npy_half_to_float(a); const npy_float bf = npy_half_to_float(b); - const npy_float outf = _basic_float_pow(af,bf); + const npy_float outf = npy_powf(af,bf); *out = npy_float_to_half(outf); } @@ -476,14 +478,10 @@ static void } /**end repeat**/ -/* - * Get the nc_powf, nc_pow, and nc_powl functions from - * the data area of the power ufunc in umathmodule. - */ - /**begin repeat * #name = cfloat, cdouble, clongdouble# * #type = npy_cfloat, npy_cdouble, npy_clongdouble# + * #c = f,,l# */ static void @name@_ctype_positive(@type@ a, @type@ *out) @@ -492,12 +490,10 @@ static void out->imag = a.imag; } -static void (*_basic_@name@_pow)(@type@ *, @type@ *, @type@ *); - static void @name@_ctype_power(@type@ a, @type@ b, @type@ *out) { - _basic_@name@_pow(&a, &b, out); + *out = npy_cpow@c@(a, b); } /**end repeat**/ @@ -570,7 +566,7 @@ static void * 1) Convert the types to the common type if both are scalars (0 return) * 2) If both are not scalars use ufunc machinery (-2 return) * 3) If both are scalars but cannot be cast to the right type - * return NotImplmented (-1 return) + * return NotImplemented (-1 return) * * 4) Perform the function on the C-type. * 5) If an error condition occurred, check to see @@ -1429,24 +1425,53 @@ static PyObject * /**begin repeat * + * #name = byte, ubyte, short, ushort, int, uint, + * long, ulong, longlong, ulonglong, + * half, float, double, longdouble, + * cfloat, cdouble, clongdouble# + * #Name = Byte, UByte, Short, UShort, Int, UInt, + * Long, ULong, LongLong, ULongLong, + * Half, Float, Double, LongDouble, + * CFloat, CDouble, CLongDouble# + * #cmplx = 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1# + * #to_ctype = , , , , , , , , , , npy_half_to_double, , , , , , # + * #func = PyFloat_FromDouble*17# + */ +static NPY_INLINE PyObject * +@name@_float(PyObject *obj) +{ +#if @cmplx@ + if (emit_complexwarning() < 0) { + return NULL; + } + return @func@(@to_ctype@(PyArrayScalar_VAL(obj, @Name@).real)); +#else + return @func@(@to_ctype@(PyArrayScalar_VAL(obj, @Name@))); +#endif +} +/**end repeat**/ + + +#if !defined(NPY_PY3K) + +/**begin repeat + * * #name = (byte, ubyte, short, ushort, int, uint, * long, ulong, longlong, ulonglong, * half, float, double, longdouble, - * cfloat, cdouble, clongdouble)*2# + * cfloat, cdouble, clongdouble)# * #Name = (Byte, UByte, Short, UShort, Int, UInt, * Long, ULong, LongLong, ULongLong, * Half, Float, Double, LongDouble, - * CFloat, CDouble, CLongDouble)*2# - * #cmplx = (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1)*2# - * #to_ctype = (, , , , , , , , , , npy_half_to_double, , , , , , )*2# - * #which = long*17, float*17# + * CFloat, CDouble, CLongDouble)# + * #cmplx = (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1)# + * #to_ctype = (, , , , , , , , , , npy_half_to_double, , , , , , )# * #func = (PyLong_FromLongLong, PyLong_FromUnsignedLongLong)*5, * PyLong_FromDouble*3, npy_longdouble_to_PyLong, - * PyLong_FromDouble*2, npy_longdouble_to_PyLong, - * PyFloat_FromDouble*17# + * PyLong_FromDouble*2, npy_longdouble_to_PyLong# */ static NPY_INLINE PyObject * -@name@_@which@(PyObject *obj) +@name@_long(PyObject *obj) { #if @cmplx@ if (emit_complexwarning() < 0) { @@ -1459,8 +1484,6 @@ static NPY_INLINE PyObject * } /**end repeat**/ -#if !defined(NPY_PY3K) - /**begin repeat * * #name = (byte, ubyte, short, ushort, int, uint, @@ -1652,52 +1675,9 @@ add_scalarmath(void) /**end repeat**/ } -static int -get_functions(PyObject * mm) -{ - PyObject *obj; - void **funcdata; - char *signatures; - int i, j; - int ret = -1; - - /* Get the nc_pow functions */ - /* Get the pow functions */ - obj = PyObject_GetAttrString(mm, "power"); - if (obj == NULL) { - goto fail; - } - funcdata = ((PyUFuncObject *)obj)->data; - signatures = ((PyUFuncObject *)obj)->types; - - i = 0; - j = 0; - while (signatures[i] != NPY_FLOAT) { - i += 3; - j++; - } - _basic_float_pow = funcdata[j]; - _basic_double_pow = funcdata[j + 1]; - _basic_longdouble_pow = funcdata[j + 2]; - _basic_cfloat_pow = funcdata[j + 3]; - _basic_cdouble_pow = funcdata[j + 4]; - _basic_clongdouble_pow = funcdata[j + 5]; - Py_DECREF(obj); - - return ret = 0; - - fail: - Py_DECREF(mm); - return ret; -} - NPY_NO_EXPORT int initscalarmath(PyObject * m) { - if (get_functions(m) < 0) { - return -1; - } - add_scalarmath(); return 0; diff --git a/numpy/core/src/umath/simd.inc.src b/numpy/core/src/umath/simd.inc.src index 9816a1da4..74f52cc9d 100644 --- a/numpy/core/src/umath/simd.inc.src +++ b/numpy/core/src/umath/simd.inc.src @@ -130,18 +130,50 @@ abs_ptrdiff(char *a, char *b) */ /**begin repeat - * #ISA = AVX2, AVX512F# - * #isa = avx2, avx512f# + * #ISA = FMA, AVX512F# + * #isa = fma, avx512f# + * #CHK = HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS# * #REGISTER_SIZE = 32, 64# */ /* prototypes */ /**begin repeat1 + * #type = npy_float, npy_double# + * #TYPE = FLOAT, DOUBLE# + */ + +/**begin repeat2 + * #func = sqrt, absolute, square, reciprocal, rint, floor, ceil, trunc# + */ + +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS +static NPY_INLINE NPY_GCC_TARGET_@ISA@ void +@ISA@_@func@_@TYPE@(@type@ *, @type@ *, const npy_intp n, const npy_intp stride); +#endif + +static NPY_INLINE int +run_unary_@isa@_@func@_@TYPE@(char **args, npy_intp *dimensions, npy_intp *steps) +{ +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS + if (IS_OUTPUT_BLOCKABLE_UNARY(sizeof(@type@), @REGISTER_SIZE@)) { + @ISA@_@func@_@TYPE@((@type@*)args[1], (@type@*)args[0], dimensions[0], steps[0]); + return 1; + } + else + return 0; +#endif + return 0; +} + +/**end repeat2**/ +/**end repeat1**/ + +/**begin repeat1 * #func = exp, log# */ -#if defined HAVE_ATTRIBUTE_TARGET_@ISA@_WITH_INTRINSICS && defined NPY_HAVE_SSE2_INTRINSICS +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS static NPY_INLINE void @ISA@_@func@_FLOAT(npy_float *, npy_float *, const npy_intp n, const npy_intp stride); #endif @@ -149,7 +181,7 @@ static NPY_INLINE void static NPY_INLINE int run_unary_@isa@_@func@_FLOAT(char **args, npy_intp *dimensions, npy_intp *steps) { -#if defined HAVE_ATTRIBUTE_TARGET_@ISA@_WITH_INTRINSICS && defined NPY_HAVE_SSE2_INTRINSICS +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS if (IS_OUTPUT_BLOCKABLE_UNARY(sizeof(npy_float), @REGISTER_SIZE@)) { @ISA@_@func@_FLOAT((npy_float*)args[1], (npy_float*)args[0], dimensions[0], steps[0]); return 1; @@ -162,8 +194,26 @@ run_unary_@isa@_@func@_FLOAT(char **args, npy_intp *dimensions, npy_intp *steps) /**end repeat1**/ -/**end repeat**/ +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS +static NPY_INLINE void +@ISA@_sincos_FLOAT(npy_float *, npy_float *, const npy_intp n, const npy_intp steps, NPY_TRIG_OP); +#endif +static NPY_INLINE int +run_unary_@isa@_sincos_FLOAT(char **args, npy_intp *dimensions, npy_intp *steps, NPY_TRIG_OP my_trig_op) +{ +#if defined @CHK@ && defined NPY_HAVE_SSE2_INTRINSICS + if (IS_OUTPUT_BLOCKABLE_UNARY(sizeof(npy_float), @REGISTER_SIZE@)) { + @ISA@_sincos_FLOAT((npy_float*)args[1], (npy_float*)args[0], dimensions[0], steps[0], my_trig_op); + return 1; + } + else + return 0; +#endif + return 0; +} + +/**end repeat**/ /**begin repeat @@ -997,7 +1047,7 @@ sse2_sqrt_@TYPE@(@type@ * op, @type@ * ip, const npy_intp n) LOOP_BLOCK_ALIGN_VAR(op, @type@, VECTOR_SIZE_BYTES) { op[i] = @scalarf@(ip[i]); } - assert(n < (VECTOR_SIZE_BYTES / sizeof(@type@)) || + assert((npy_uintp)n < (VECTOR_SIZE_BYTES / sizeof(@type@)) || npy_is_aligned(&op[i], VECTOR_SIZE_BYTES)); if (npy_is_aligned(&ip[i], VECTOR_SIZE_BYTES)) { LOOP_BLOCKED(@type@, VECTOR_SIZE_BYTES) { @@ -1049,7 +1099,7 @@ sse2_@kind@_@TYPE@(@type@ * op, @type@ * ip, const npy_intp n) LOOP_BLOCK_ALIGN_VAR(op, @type@, VECTOR_SIZE_BYTES) { op[i] = @scalar@_@type@(ip[i]); } - assert(n < (VECTOR_SIZE_BYTES / sizeof(@type@)) || + assert((npy_uintp)n < (VECTOR_SIZE_BYTES / sizeof(@type@)) || npy_is_aligned(&op[i], VECTOR_SIZE_BYTES)); if (npy_is_aligned(&ip[i], VECTOR_SIZE_BYTES)) { LOOP_BLOCKED(@type@, VECTOR_SIZE_BYTES) { @@ -1084,7 +1134,7 @@ sse2_@kind@_@TYPE@(@type@ * ip, @type@ * op, const npy_intp n) /* Order of operations important for MSVC 2015 */ *op = (*op @OP@ ip[i] || npy_isnan(*op)) ? *op : ip[i]; } - assert(n < (stride) || npy_is_aligned(&ip[i], VECTOR_SIZE_BYTES)); + assert((npy_uintp)n < (stride) || npy_is_aligned(&ip[i], VECTOR_SIZE_BYTES)); if (i + 3 * stride <= n) { /* load the first elements */ @vtype@ c1 = @vpre@_load_@vsuf@((@type@*)&ip[i]); @@ -1123,56 +1173,110 @@ sse2_@kind@_@TYPE@(@type@ * ip, @type@ * op, const npy_intp n) /* bunch of helper functions used in ISA_exp/log_FLOAT*/ #if defined HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_fmadd(__m256 a, __m256 b, __m256 c) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_get_full_load_mask_ps(void) { - return _mm256_add_ps(_mm256_mul_ps(a, b), c); + return _mm256_set1_ps(-1.0); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_get_full_load_mask(void) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256i +fma_get_full_load_mask_pd(void) { - return _mm256_set1_ps(-1.0); + return _mm256_castpd_si256(_mm256_set1_pd(-1.0)); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_get_partial_load_mask(const npy_int num_lanes, const npy_int total_elem) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_get_partial_load_mask_ps(const npy_int num_elem, const npy_int num_lanes) { float maskint[16] = {-1.0,-1.0,-1.0,-1.0,-1.0,-1.0,-1.0,-1.0, 1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0}; - float* addr = maskint + total_elem - num_lanes; + float* addr = maskint + num_lanes - num_elem; return _mm256_loadu_ps(addr); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_masked_gather(__m256 src, - npy_float* addr, - __m256i vindex, - __m256 mask) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256i +fma_get_partial_load_mask_pd(const npy_int num_elem, const npy_int num_lanes) +{ + npy_int maskint[16] = {-1,-1,-1,-1,-1,-1,-1,-1,1,1,1,1,1,1,1,1}; + npy_int* addr = maskint + 2*num_lanes - 2*num_elem; + return _mm256_loadu_si256((__m256i*) addr); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_masked_gather_ps(__m256 src, + npy_float* addr, + __m256i vindex, + __m256 mask) { return _mm256_mask_i32gather_ps(src, addr, vindex, mask, 4); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_masked_load(__m256 mask, npy_float* addr) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256d +fma_masked_gather_pd(__m256d src, + npy_double* addr, + __m128i vindex, + __m256d mask) +{ + return _mm256_mask_i32gather_pd(src, addr, vindex, mask, 8); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_masked_load_ps(__m256 mask, npy_float* addr) { return _mm256_maskload_ps(addr, _mm256_cvtps_epi32(mask)); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_set_masked_lanes(__m256 x, __m256 val, __m256 mask) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256d +fma_masked_load_pd(__m256i mask, npy_double* addr) +{ + return _mm256_maskload_pd(addr, mask); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_set_masked_lanes_ps(__m256 x, __m256 val, __m256 mask) { return _mm256_blendv_ps(x, val, mask); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_blend(__m256 x, __m256 y, __m256 ymask) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256d +fma_set_masked_lanes_pd(__m256d x, __m256d val, __m256d mask) +{ + return _mm256_blendv_pd(x, val, mask); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_blend(__m256 x, __m256 y, __m256 ymask) { return _mm256_blendv_ps(x, y, ymask); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_get_exponent(__m256 x) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_invert_mask_ps(__m256 ymask) +{ + return _mm256_andnot_ps(ymask, _mm256_set1_ps(-1.0)); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256i +fma_invert_mask_pd(__m256i ymask) +{ + return _mm256_andnot_si256(ymask, _mm256_set1_epi32(0xFFFFFFFF)); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_should_calculate_sine(__m256i k, __m256i andop, __m256i cmp) +{ + return _mm256_cvtepi32_ps( + _mm256_cmpeq_epi32(_mm256_and_si256(k, andop), cmp)); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_should_negate(__m256i k, __m256i andop, __m256i cmp) +{ + return fma_should_calculate_sine(k, andop, cmp); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_get_exponent(__m256 x) { /* * Special handling of denormals: @@ -1198,8 +1302,8 @@ avx2_get_exponent(__m256 x) return _mm256_blendv_ps(exp, denorm_exp, denormal_mask); } -static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_get_mantissa(__m256 x) +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA __m256 +fma_get_mantissa(__m256 x) { /* * Special handling of denormals: @@ -1225,7 +1329,7 @@ avx2_get_mantissa(__m256 x) } static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX2 __m256 -avx2_scalef_ps(__m256 poly, __m256 quadrant) +fma_scalef_ps(__m256 poly, __m256 quadrant) { /* * Handle denormals (which occur when quadrant <= -125): @@ -1263,48 +1367,145 @@ avx2_scalef_ps(__m256 poly, __m256 quadrant) } } +/**begin repeat + * #vsub = ps, pd# + * #vtype = __m256, __m256d# + */ +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_abs_@vsub@(@vtype@ x) +{ + return _mm256_andnot_@vsub@(_mm256_set1_@vsub@(-0.0), x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_reciprocal_@vsub@(@vtype@ x) +{ + return _mm256_div_@vsub@(_mm256_set1_@vsub@(1.0f), x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_rint_@vsub@(@vtype@ x) +{ + return _mm256_round_@vsub@(x, _MM_FROUND_TO_NEAREST_INT); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_floor_@vsub@(@vtype@ x) +{ + return _mm256_round_@vsub@(x, _MM_FROUND_TO_NEG_INF); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_ceil_@vsub@(@vtype@ x) +{ + return _mm256_round_@vsub@(x, _MM_FROUND_TO_POS_INF); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_FMA @vtype@ +fma_trunc_@vsub@(@vtype@ x) +{ + return _mm256_round_@vsub@(x, _MM_FROUND_TO_ZERO); +} +/**end repeat**/ #endif #if defined HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask16 -avx512_get_full_load_mask(void) +avx512_get_full_load_mask_ps(void) { return 0xFFFF; } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask8 +avx512_get_full_load_mask_pd(void) +{ + return 0xFF; +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask16 -avx512_get_partial_load_mask(const npy_int num_elem, const npy_int total_elem) +avx512_get_partial_load_mask_ps(const npy_int num_elem, const npy_int total_elem) { return (0x0001 << num_elem) - 0x0001; } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask8 +avx512_get_partial_load_mask_pd(const npy_int num_elem, const npy_int total_elem) +{ + return (0x01 << num_elem) - 0x01; +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512 -avx512_masked_gather(__m512 src, - npy_float* addr, - __m512i vindex, - __mmask16 kmask) +avx512_masked_gather_ps(__m512 src, + npy_float* addr, + __m512i vindex, + __mmask16 kmask) { return _mm512_mask_i32gather_ps(src, kmask, vindex, addr, 4); } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512d +avx512_masked_gather_pd(__m512d src, + npy_double* addr, + __m256i vindex, + __mmask8 kmask) +{ + return _mm512_mask_i32gather_pd(src, kmask, vindex, addr, 8); +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512 -avx512_masked_load(__mmask16 mask, npy_float* addr) +avx512_masked_load_ps(__mmask16 mask, npy_float* addr) { return _mm512_maskz_loadu_ps(mask, (__m512 *)addr); } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512d +avx512_masked_load_pd(__mmask8 mask, npy_double* addr) +{ + return _mm512_maskz_loadu_pd(mask, (__m512d *)addr); +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512 -avx512_set_masked_lanes(__m512 x, __m512 val, __mmask16 mask) +avx512_set_masked_lanes_ps(__m512 x, __m512 val, __mmask16 mask) { return _mm512_mask_blend_ps(mask, x, val); } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512d +avx512_set_masked_lanes_pd(__m512d x, __m512d val, __mmask8 mask) +{ + return _mm512_mask_blend_pd(mask, x, val); +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512 avx512_blend(__m512 x, __m512 y, __mmask16 ymask) { return _mm512_mask_mov_ps(x, ymask, y); } +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask16 +avx512_invert_mask_ps(__mmask16 ymask) +{ + return _mm512_knot(ymask); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask8 +avx512_invert_mask_pd(__mmask8 ymask) +{ + return _mm512_knot(ymask); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask16 +avx512_should_calculate_sine(__m512i k, __m512i andop, __m512i cmp) +{ + return _mm512_cmpeq_epi32_mask(_mm512_and_epi32(k, andop), cmp); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __mmask16 +avx512_should_negate(__m512i k, __m512i andop, __m512i cmp) +{ + return avx512_should_calculate_sine(k, andop, cmp); +} + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F __m512 avx512_get_exponent(__m512 x) { @@ -1322,20 +1523,74 @@ avx512_scalef_ps(__m512 poly, __m512 quadrant) { return _mm512_scalef_ps(poly, quadrant); } +/**begin repeat + * #vsub = ps, pd# + * #epi_vsub = epi32, epi64# + * #vtype = __m512, __m512d# + * #and_const = 0x7fffffff, 0x7fffffffffffffffLL# + */ +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_abs_@vsub@(@vtype@ x) +{ + return (@vtype@) _mm512_and_@epi_vsub@((__m512i) x, + _mm512_set1_@epi_vsub@ (@and_const@)); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_reciprocal_@vsub@(@vtype@ x) +{ + return _mm512_div_@vsub@(_mm512_set1_@vsub@(1.0f), x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_rint_@vsub@(@vtype@ x) +{ + return _mm512_roundscale_@vsub@(x, 0x08); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_floor_@vsub@(@vtype@ x) +{ + return _mm512_roundscale_@vsub@(x, 0x09); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_ceil_@vsub@(@vtype@ x) +{ + return _mm512_roundscale_@vsub@(x, 0x0A); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_AVX512F @vtype@ +avx512_trunc_@vsub@(@vtype@ x) +{ + return _mm512_roundscale_@vsub@(x, 0x0B); +} +/**end repeat**/ #endif /**begin repeat - * #ISA = AVX2, AVX512F# - * #isa = avx2, avx512# + * #ISA = FMA, AVX512F# + * #isa = fma, avx512# * #vtype = __m256, __m512# * #vsize = 256, 512# * #or = or_ps, kor# * #vsub = , _mask# * #mask = __m256, __mmask16# - * #fmadd = avx2_fmadd,_mm512_fmadd_ps# + * #fmadd = _mm256_fmadd_ps, _mm512_fmadd_ps# + * #CHK = HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS# **/ -#if defined HAVE_ATTRIBUTE_TARGET_@ISA@_WITH_INTRINSICS +#if defined @CHK@ + +/* + * Vectorized Cody-Waite range reduction technique + * Performs the reduction step x* = x - y*C in three steps: + * 1) x* = x - y*c1 + * 2) x* = x - y*c2 + * 3) x* = x - y*c3 + * c1, c2 are exact floating points, c3 = C - c1 - c2 simulates higher precision + */ + static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ @isa@_range_reduction(@vtype@ x, @vtype@ y, @vtype@ c1, @vtype@ c2, @vtype@ c3) { @@ -1344,12 +1599,236 @@ static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ reduced_x = @fmadd@(y, c3, reduced_x); return reduced_x; } + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @mask@ +@isa@_in_range_mask(@vtype@ x, npy_float fmax, npy_float fmin) +{ + @mask@ m1 = _mm@vsize@_cmp_ps@vsub@( + x, _mm@vsize@_set1_ps(fmax), _CMP_GT_OQ); + @mask@ m2 = _mm@vsize@_cmp_ps@vsub@( + x, _mm@vsize@_set1_ps(fmin), _CMP_LT_OQ); + return _mm@vsize@_@or@(m1,m2); +} + +/* + * Approximate cosine algorithm for x \in [-PI/4, PI/4] + * Maximum ULP across all 32-bit floats = 0.875 + */ + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ +@isa@_cosine(@vtype@ x2, @vtype@ invf8, @vtype@ invf6, @vtype@ invf4, + @vtype@ invf2, @vtype@ invf0) +{ + @vtype@ cos = @fmadd@(invf8, x2, invf6); + cos = @fmadd@(cos, x2, invf4); + cos = @fmadd@(cos, x2, invf2); + cos = @fmadd@(cos, x2, invf0); + return cos; +} + +/* + * Approximate sine algorithm for x \in [-PI/4, PI/4] + * Maximum ULP across all 32-bit floats = 0.647 + */ + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ +@isa@_sine(@vtype@ x, @vtype@ x2, @vtype@ invf9, @vtype@ invf7, + @vtype@ invf5, @vtype@ invf3, + @vtype@ zero) +{ + @vtype@ sin = @fmadd@(invf9, x2, invf7); + sin = @fmadd@(sin, x2, invf5); + sin = @fmadd@(sin, x2, invf3); + sin = @fmadd@(sin, x2, zero); + sin = @fmadd@(sin, x, x); + return sin; +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ +@isa@_sqrt_ps(@vtype@ x) +{ + return _mm@vsize@_sqrt_ps(x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@d +@isa@_sqrt_pd(@vtype@d x) +{ + return _mm@vsize@_sqrt_pd(x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ +@isa@_square_ps(@vtype@ x) +{ + return _mm@vsize@_mul_ps(x,x); +} + +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@d +@isa@_square_pd(@vtype@d x) +{ + return _mm@vsize@_mul_pd(x,x); +} + +#endif +/**end repeat**/ + + +/**begin repeat + * #ISA = FMA, AVX512F# + * #isa = fma, avx512# + * #vsize = 256, 512# + * #BYTES = 32, 64# + * #cvtps_epi32 = _mm256_cvtps_epi32, # + * #mask = __m256, __mmask16# + * #vsub = , _mask# + * #vtype = __m256, __m512# + * #cvtps_epi32 = _mm256_cvtps_epi32, # + * #masked_store = _mm256_maskstore_ps, _mm512_mask_storeu_ps# + * #CHK = HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS# + */ + +/**begin repeat1 + * #func = sqrt, absolute, square, reciprocal, rint, ceil, floor, trunc# + * #vectorf = sqrt, abs, square, reciprocal, rint, ceil, floor, trunc# + * #replace_0_with_1 = 0, 0, 0, 1, 0, 0, 0, 0# + */ + +#if defined @CHK@ +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void +@ISA@_@func@_FLOAT(npy_float* op, + npy_float* ip, + const npy_intp array_size, + const npy_intp steps) +{ + const npy_intp stride = steps/sizeof(npy_float); + const npy_int num_lanes = @BYTES@/sizeof(npy_float); + npy_intp num_remaining_elements = array_size; + @vtype@ ones_f = _mm@vsize@_set1_ps(1.0f); + @mask@ load_mask = @isa@_get_full_load_mask_ps(); +#if @replace_0_with_1@ + @mask@ inv_load_mask = @isa@_invert_mask_ps(load_mask); +#endif + npy_int indexarr[16]; + for (npy_int ii = 0; ii < 16; ii++) { + indexarr[ii] = ii*stride; + } + @vtype@i vindex = _mm@vsize@_loadu_si@vsize@((@vtype@i*)&indexarr[0]); + + while (num_remaining_elements > 0) { + if (num_remaining_elements < num_lanes) { + load_mask = @isa@_get_partial_load_mask_ps(num_remaining_elements, + num_lanes); +#if @replace_0_with_1@ + inv_load_mask = @isa@_invert_mask_ps(load_mask); +#endif + } + @vtype@ x; + if (stride == 1) { + x = @isa@_masked_load_ps(load_mask, ip); +#if @replace_0_with_1@ + /* + * Replace masked elements with 1.0f to avoid divide by zero fp + * exception in reciprocal + */ + x = @isa@_set_masked_lanes_ps(x, ones_f, inv_load_mask); +#endif + } + else { + x = @isa@_masked_gather_ps(ones_f, ip, vindex, load_mask); + } + @vtype@ out = @isa@_@vectorf@_ps(x); + @masked_store@(op, @cvtps_epi32@(load_mask), out); + + ip += num_lanes*stride; + op += num_lanes; + num_remaining_elements -= num_lanes; + } +} +#endif +/**end repeat1**/ +/**end repeat**/ + +/**begin repeat + * #ISA = FMA, AVX512F# + * #isa = fma, avx512# + * #vsize = 256, 512# + * #BYTES = 32, 64# + * #cvtps_epi32 = _mm256_cvtps_epi32, # + * #mask = __m256i, __mmask8# + * #vsub = , _mask# + * #vtype = __m256d, __m512d# + * #vindextype = __m128i, __m256i# + * #vindexsize = 128, 256# + * #vindexload = _mm_loadu_si128, _mm256_loadu_si256# + * #cvtps_epi32 = _mm256_cvtpd_epi32, # + * #castmask = _mm256_castsi256_pd, # + * #masked_store = _mm256_maskstore_pd, _mm512_mask_storeu_pd# + * #CHK = HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS# + */ + +/**begin repeat1 + * #func = sqrt, absolute, square, reciprocal, rint, ceil, floor, trunc# + * #vectorf = sqrt, abs, square, reciprocal, rint, ceil, floor, trunc# + * #replace_0_with_1 = 0, 0, 0, 1, 0, 0, 0, 0# + */ + +#if defined @CHK@ +static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void +@ISA@_@func@_DOUBLE(npy_double* op, + npy_double* ip, + const npy_intp array_size, + const npy_intp steps) +{ + const npy_intp stride = steps/sizeof(npy_double); + const npy_int num_lanes = @BYTES@/sizeof(npy_double); + npy_intp num_remaining_elements = array_size; + @mask@ load_mask = @isa@_get_full_load_mask_pd(); +#if @replace_0_with_1@ + @mask@ inv_load_mask = @isa@_invert_mask_pd(load_mask); +#endif + @vtype@ ones_d = _mm@vsize@_set1_pd(1.0f); + npy_int indexarr[8]; + for (npy_int ii = 0; ii < 8; ii++) { + indexarr[ii] = ii*stride; + } + @vindextype@ vindex = @vindexload@((@vindextype@*)&indexarr[0]); + + while (num_remaining_elements > 0) { + if (num_remaining_elements < num_lanes) { + load_mask = @isa@_get_partial_load_mask_pd(num_remaining_elements, + num_lanes); +#if @replace_0_with_1@ + inv_load_mask = @isa@_invert_mask_pd(load_mask); +#endif + } + @vtype@ x; + if (stride == 1) { + x = @isa@_masked_load_pd(load_mask, ip); +#if @replace_0_with_1@ + /* + * Replace masked elements with 1.0f to avoid divide by zero fp + * exception in reciprocal + */ + x = @isa@_set_masked_lanes_pd(x, ones_d, @castmask@(inv_load_mask)); #endif + } + else { + x = @isa@_masked_gather_pd(ones_d, ip, vindex, @castmask@(load_mask)); + } + @vtype@ out = @isa@_@vectorf@_pd(x); + @masked_store@(op, load_mask, out); + + ip += num_lanes*stride; + op += num_lanes; + num_remaining_elements -= num_lanes; + } +} +#endif +/**end repeat1**/ /**end repeat**/ /**begin repeat - * #ISA = AVX2, AVX512F# - * #isa = avx2, avx512# + * #ISA = FMA, AVX512F# + * #isa = fma, avx512# * #vtype = __m256, __m512# * #vsize = 256, 512# * #BYTES = 32, 64# @@ -1358,13 +1837,164 @@ static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ * #or_masks =_mm256_or_ps, _mm512_kor# * #and_masks =_mm256_and_ps, _mm512_kand# * #xor_masks =_mm256_xor_ps, _mm512_kxor# - * #fmadd = avx2_fmadd,_mm512_fmadd_ps# + * #fmadd = _mm256_fmadd_ps, _mm512_fmadd_ps# * #mask_to_int = _mm256_movemask_ps, # * #full_mask= 0xFF, 0xFFFF# * #masked_store = _mm256_maskstore_ps, _mm512_mask_storeu_ps# * #cvtps_epi32 = _mm256_cvtps_epi32, # + * #CHK = HAVE_ATTRIBUTE_TARGET_AVX2_WITH_INTRINSICS, HAVE_ATTRIBUTE_TARGET_AVX512F_WITH_INTRINSICS# */ +/* + * Vectorized approximate sine/cosine algorithms: The following code is a + * vectorized version of the algorithm presented here: + * https://stackoverflow.com/questions/30463616/payne-hanek-algorithm-implementation-in-c/30465751#30465751 + * (1) Load data in ZMM/YMM registers and generate mask for elements that are + * within range [-71476.0625f, 71476.0625f] for cosine and [-117435.992f, + * 117435.992f] for sine. + * (2) For elements within range, perform range reduction using Cody-Waite's + * method: x* = x - y*PI/2, where y = rint(x*2/PI). x* \in [-PI/4, PI/4]. + * (3) Map cos(x) to (+/-)sine or (+/-)cosine of x* based on the quadrant k = + * int(y). + * (4) For elements outside that range, Cody-Waite reduction peforms poorly + * leading to catastrophic cancellation. We compute cosine by calling glibc in + * a scalar fashion. + * (5) Vectorized implementation has a max ULP of 1.49 and performs at least + * 5-7x faster than scalar implementations when magnitude of all elements in + * the array < 71476.0625f (117435.992f for sine). Worst case performance is + * when all the elements are large leading to about 1-2% reduction in + * performance. + */ + +#if defined @CHK@ +static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void +@ISA@_sincos_FLOAT(npy_float * op, + npy_float * ip, + const npy_intp array_size, + const npy_intp steps, + NPY_TRIG_OP my_trig_op) +{ + const npy_intp stride = steps/sizeof(npy_float); + const npy_int num_lanes = @BYTES@/sizeof(npy_float); + npy_float large_number = 71476.0625f; + if (my_trig_op == npy_compute_sin) { + large_number = 117435.992f; + } + + /* Load up frequently used constants */ + @vtype@i zeros = _mm@vsize@_set1_epi32(0); + @vtype@i ones = _mm@vsize@_set1_epi32(1); + @vtype@i twos = _mm@vsize@_set1_epi32(2); + @vtype@ two_over_pi = _mm@vsize@_set1_ps(NPY_TWO_O_PIf); + @vtype@ codyw_c1 = _mm@vsize@_set1_ps(NPY_CODY_WAITE_PI_O_2_HIGHf); + @vtype@ codyw_c2 = _mm@vsize@_set1_ps(NPY_CODY_WAITE_PI_O_2_MEDf); + @vtype@ codyw_c3 = _mm@vsize@_set1_ps(NPY_CODY_WAITE_PI_O_2_LOWf); + @vtype@ cos_invf0 = _mm@vsize@_set1_ps(NPY_COEFF_INVF0_COSINEf); + @vtype@ cos_invf2 = _mm@vsize@_set1_ps(NPY_COEFF_INVF2_COSINEf); + @vtype@ cos_invf4 = _mm@vsize@_set1_ps(NPY_COEFF_INVF4_COSINEf); + @vtype@ cos_invf6 = _mm@vsize@_set1_ps(NPY_COEFF_INVF6_COSINEf); + @vtype@ cos_invf8 = _mm@vsize@_set1_ps(NPY_COEFF_INVF8_COSINEf); + @vtype@ sin_invf3 = _mm@vsize@_set1_ps(NPY_COEFF_INVF3_SINEf); + @vtype@ sin_invf5 = _mm@vsize@_set1_ps(NPY_COEFF_INVF5_SINEf); + @vtype@ sin_invf7 = _mm@vsize@_set1_ps(NPY_COEFF_INVF7_SINEf); + @vtype@ sin_invf9 = _mm@vsize@_set1_ps(NPY_COEFF_INVF9_SINEf); + @vtype@ cvt_magic = _mm@vsize@_set1_ps(NPY_RINT_CVT_MAGICf); + @vtype@ zero_f = _mm@vsize@_set1_ps(0.0f); + @vtype@ quadrant, reduced_x, reduced_x2, cos, sin; + @vtype@i iquadrant; + @mask@ nan_mask, glibc_mask, sine_mask, negate_mask; + @mask@ load_mask = @isa@_get_full_load_mask_ps(); + npy_intp num_remaining_elements = array_size; + npy_int indexarr[16]; + for (npy_int ii = 0; ii < 16; ii++) { + indexarr[ii] = ii*stride; + } + @vtype@i vindex = _mm@vsize@_loadu_si@vsize@((@vtype@i*)&indexarr[0]); + + while (num_remaining_elements > 0) { + + if (num_remaining_elements < num_lanes) { + load_mask = @isa@_get_partial_load_mask_ps(num_remaining_elements, + num_lanes); + } + + @vtype@ x; + if (stride == 1) { + x = @isa@_masked_load_ps(load_mask, ip); + } + else { + x = @isa@_masked_gather_ps(zero_f, ip, vindex, load_mask); + } + + /* + * For elements outside of this range, Cody-Waite's range reduction + * becomes inaccurate and we will call glibc to compute cosine for + * these numbers + */ + + glibc_mask = @isa@_in_range_mask(x, large_number,-large_number); + glibc_mask = @and_masks@(load_mask, glibc_mask); + nan_mask = _mm@vsize@_cmp_ps@vsub@(x, x, _CMP_NEQ_UQ); + x = @isa@_set_masked_lanes_ps(x, zero_f, @or_masks@(nan_mask, glibc_mask)); + npy_int iglibc_mask = @mask_to_int@(glibc_mask); + + if (iglibc_mask != @full_mask@) { + quadrant = _mm@vsize@_mul_ps(x, two_over_pi); + + /* round to nearest */ + quadrant = _mm@vsize@_add_ps(quadrant, cvt_magic); + quadrant = _mm@vsize@_sub_ps(quadrant, cvt_magic); + + /* Cody-Waite's range reduction algorithm */ + reduced_x = @isa@_range_reduction(x, quadrant, + codyw_c1, codyw_c2, codyw_c3); + reduced_x2 = _mm@vsize@_mul_ps(reduced_x, reduced_x); + + /* compute cosine and sine */ + cos = @isa@_cosine(reduced_x2, cos_invf8, cos_invf6, cos_invf4, + cos_invf2, cos_invf0); + sin = @isa@_sine(reduced_x, reduced_x2, sin_invf9, sin_invf7, + sin_invf5, sin_invf3, zero_f); + + iquadrant = _mm@vsize@_cvtps_epi32(quadrant); + if (my_trig_op == npy_compute_cos) { + iquadrant = _mm@vsize@_add_epi32(iquadrant, ones); + } + + /* blend sin and cos based on the quadrant */ + sine_mask = @isa@_should_calculate_sine(iquadrant, ones, zeros); + cos = @isa@_blend(cos, sin, sine_mask); + + /* multiply by -1 for appropriate elements */ + negate_mask = @isa@_should_negate(iquadrant, twos, twos); + cos = @isa@_blend(cos, _mm@vsize@_sub_ps(zero_f, cos), negate_mask); + cos = @isa@_set_masked_lanes_ps(cos, _mm@vsize@_set1_ps(NPY_NANF), nan_mask); + + @masked_store@(op, @cvtps_epi32@(load_mask), cos); + } + + /* process elements using glibc for large elements */ + if (my_trig_op == npy_compute_cos) { + for (int ii = 0; iglibc_mask != 0; ii++) { + if (iglibc_mask & 0x01) { + op[ii] = npy_cosf(ip[ii]); + } + iglibc_mask = iglibc_mask >> 1; + } + } + else { + for (int ii = 0; iglibc_mask != 0; ii++) { + if (iglibc_mask & 0x01) { + op[ii] = npy_sinf(ip[ii]); + } + iglibc_mask = iglibc_mask >> 1; + } + } + ip += num_lanes*stride; + op += num_lanes; + num_remaining_elements -= num_lanes; + } +} /* * Vectorized implementation of exp using AVX2 and AVX512: @@ -1381,7 +2011,6 @@ static NPY_INLINE NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ @vtype@ * same x = 0xc2781e37) */ -#if defined HAVE_ATTRIBUTE_TARGET_@ISA@_WITH_INTRINSICS static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void @ISA@_exp_FLOAT(npy_float * op, npy_float * ip, @@ -1417,27 +2046,27 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void @vtype@i vindex = _mm@vsize@_loadu_si@vsize@((@vtype@i*)&indexarr[0]); @mask@ xmax_mask, xmin_mask, nan_mask, inf_mask; - @mask@ overflow_mask = @isa@_get_partial_load_mask(0, num_lanes); - @mask@ load_mask = @isa@_get_full_load_mask(); + @mask@ overflow_mask = @isa@_get_partial_load_mask_ps(0, num_lanes); + @mask@ load_mask = @isa@_get_full_load_mask_ps(); npy_intp num_remaining_elements = array_size; while (num_remaining_elements > 0) { if (num_remaining_elements < num_lanes) { - load_mask = @isa@_get_partial_load_mask(num_remaining_elements, - num_lanes); + load_mask = @isa@_get_partial_load_mask_ps(num_remaining_elements, + num_lanes); } @vtype@ x; if (stride == 1) { - x = @isa@_masked_load(load_mask, ip); + x = @isa@_masked_load_ps(load_mask, ip); } else { - x = @isa@_masked_gather(zeros_f, ip, vindex, load_mask); + x = @isa@_masked_gather_ps(zeros_f, ip, vindex, load_mask); } nan_mask = _mm@vsize@_cmp_ps@vsub@(x, x, _CMP_NEQ_UQ); - x = @isa@_set_masked_lanes(x, zeros_f, nan_mask); + x = @isa@_set_masked_lanes_ps(x, zeros_f, nan_mask); xmax_mask = _mm@vsize@_cmp_ps@vsub@(x, _mm@vsize@_set1_ps(xmax), _CMP_GE_OQ); xmin_mask = _mm@vsize@_cmp_ps@vsub@(x, _mm@vsize@_set1_ps(xmin), _CMP_LE_OQ); @@ -1445,7 +2074,7 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void overflow_mask = @or_masks@(overflow_mask, @xor_masks@(xmax_mask, inf_mask)); - x = @isa@_set_masked_lanes(x, zeros_f, @or_masks@( + x = @isa@_set_masked_lanes_ps(x, zeros_f, @or_masks@( @or_masks@(nan_mask, xmin_mask), xmax_mask)); quadrant = _mm@vsize@_mul_ps(x, log2e); @@ -1478,9 +2107,9 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void * elem < xmin; return 0.0f * elem = +/- nan, return nan */ - poly = @isa@_set_masked_lanes(poly, _mm@vsize@_set1_ps(NPY_NANF), nan_mask); - poly = @isa@_set_masked_lanes(poly, inf, xmax_mask); - poly = @isa@_set_masked_lanes(poly, zeros_f, xmin_mask); + poly = @isa@_set_masked_lanes_ps(poly, _mm@vsize@_set1_ps(NPY_NANF), nan_mask); + poly = @isa@_set_masked_lanes_ps(poly, inf, xmax_mask); + poly = @isa@_set_masked_lanes_ps(poly, zeros_f, xmin_mask); @masked_store@(op, @cvtps_epi32@(load_mask), poly); @@ -1545,24 +2174,24 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void @vtype@ poly, num_poly, denom_poly, exponent; @mask@ inf_mask, nan_mask, sqrt2_mask, zero_mask, negx_mask; - @mask@ invalid_mask = @isa@_get_partial_load_mask(0, num_lanes); + @mask@ invalid_mask = @isa@_get_partial_load_mask_ps(0, num_lanes); @mask@ divide_by_zero_mask = invalid_mask; - @mask@ load_mask = @isa@_get_full_load_mask(); + @mask@ load_mask = @isa@_get_full_load_mask_ps(); npy_intp num_remaining_elements = array_size; while (num_remaining_elements > 0) { if (num_remaining_elements < num_lanes) { - load_mask = @isa@_get_partial_load_mask(num_remaining_elements, - num_lanes); + load_mask = @isa@_get_partial_load_mask_ps(num_remaining_elements, + num_lanes); } @vtype@ x_in; if (stride == 1) { - x_in = @isa@_masked_load(load_mask, ip); + x_in = @isa@_masked_load_ps(load_mask, ip); } else { - x_in = @isa@_masked_gather(zeros_f, ip, vindex, load_mask); + x_in = @isa@_masked_gather_ps(zeros_f, ip, vindex, load_mask); } negx_mask = _mm@vsize@_cmp_ps@vsub@(x_in, zeros_f, _CMP_LT_OQ); @@ -1573,7 +2202,7 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void @and_masks@(zero_mask, load_mask)); invalid_mask = @or_masks@(invalid_mask, negx_mask); - @vtype@ x = @isa@_set_masked_lanes(x_in, zeros_f, negx_mask); + @vtype@ x = @isa@_set_masked_lanes_ps(x_in, zeros_f, negx_mask); /* set x = normalized mantissa */ exponent = @isa@_get_exponent(x); @@ -1607,10 +2236,10 @@ static NPY_GCC_OPT_3 NPY_GCC_TARGET_@ISA@ void * x = +/- NAN; return NAN * x = 0.0f; return -INF */ - poly = @isa@_set_masked_lanes(poly, nan, nan_mask); - poly = @isa@_set_masked_lanes(poly, neg_nan, negx_mask); - poly = @isa@_set_masked_lanes(poly, neg_inf, zero_mask); - poly = @isa@_set_masked_lanes(poly, inf, inf_mask); + poly = @isa@_set_masked_lanes_ps(poly, nan, nan_mask); + poly = @isa@_set_masked_lanes_ps(poly, neg_nan, negx_mask); + poly = @isa@_set_masked_lanes_ps(poly, neg_inf, zero_mask); + poly = @isa@_set_masked_lanes_ps(poly, inf, inf_mask); @masked_store@(op, @cvtps_epi32@(load_mask), poly); diff --git a/numpy/core/src/umath/ufunc_object.c b/numpy/core/src/umath/ufunc_object.c index 174703fb1..1dc581977 100644 --- a/numpy/core/src/umath/ufunc_object.c +++ b/numpy/core/src/umath/ufunc_object.c @@ -1193,34 +1193,11 @@ get_ufunc_arguments(PyUFuncObject *ufunc, } } else { - /* - * If the deprecated behavior is ever removed, - * keep only the else branch of this if-else - */ - if (PyArray_Check(out_kwd) || out_kwd == Py_None) { - if (DEPRECATE("passing a single array to the " - "'out' keyword argument of a " - "ufunc with\n" - "more than one output will " - "result in an error in the " - "future") < 0) { - /* The future error message */ - PyErr_SetString(PyExc_TypeError, - "'out' must be a tuple of arrays"); - goto fail; - } - if (_set_out_array(out_kwd, out_op+nin) < 0) { - goto fail; - } - } - else { - PyErr_SetString(PyExc_TypeError, - nout > 1 ? "'out' must be a tuple " - "of arrays" : - "'out' must be an array or a " - "tuple of a single array"); - goto fail; - } + PyErr_SetString(PyExc_TypeError, + nout > 1 ? "'out' must be a tuple of arrays" : + "'out' must be an array or a tuple with " + "a single array"); + goto fail; } } /* @@ -2297,7 +2274,7 @@ _parse_axes_arg(PyUFuncObject *ufunc, int op_core_num_dims[], PyObject *axes, * Returns 0 on success, and -1 on failure */ static int -_parse_axis_arg(PyUFuncObject *ufunc, int core_num_dims[], PyObject *axis, +_parse_axis_arg(PyUFuncObject *ufunc, const int core_num_dims[], PyObject *axis, PyArrayObject **op, int broadcast_ndim, int **remap_axis) { int nop = ufunc->nargs; int iop, axis_int; @@ -2368,7 +2345,7 @@ _parse_axis_arg(PyUFuncObject *ufunc, int core_num_dims[], PyObject *axis, */ static int _get_coredim_sizes(PyUFuncObject *ufunc, PyArrayObject **op, - int *op_core_num_dims, npy_uint32 *core_dim_flags, + const int *op_core_num_dims, npy_uint32 *core_dim_flags, npy_intp *core_dim_sizes, int **remap_axis) { int i; int nin = ufunc->nin; @@ -4081,8 +4058,8 @@ PyUFunc_Reduceat(PyUFuncObject *ufunc, PyArrayObject *arr, PyArrayObject *ind, for (i = 0; i < ind_size; ++i) { if (reduceat_ind[i] < 0 || reduceat_ind[i] >= red_axis_size) { PyErr_Format(PyExc_IndexError, - "index %d out-of-bounds in %s.%s [0, %d)", - (int)reduceat_ind[i], ufunc_name, opname, (int)red_axis_size); + "index %" NPY_INTP_FMT " out-of-bounds in %s.%s [0, %" NPY_INTP_FMT ")", + reduceat_ind[i], ufunc_name, opname, red_axis_size); return NULL; } } @@ -4766,6 +4743,9 @@ ufunc_generic_call(PyUFuncObject *ufunc, PyObject *args, PyObject *kwds) wrapped = _apply_array_wrap(wraparr[i], mps[j], &context); mps[j] = NULL; /* Prevent fail double-freeing this */ if (wrapped == NULL) { + for (j = 0; j < i; j++) { + Py_DECREF(retobj[j]); + } goto fail; } @@ -4863,7 +4843,7 @@ ufunc_seterr(PyObject *NPY_UNUSED(dummy), PyObject *args) NPY_NO_EXPORT int PyUFunc_ReplaceLoopBySignature(PyUFuncObject *func, PyUFuncGenericFunction newfunc, - int *signature, + const int *signature, PyUFuncGenericFunction *oldfunc) { int i, j; @@ -4918,7 +4898,7 @@ PyUFunc_FromFuncAndDataAndSignatureAndIdentity(PyUFuncGenericFunction *func, voi char *types, int ntypes, int nin, int nout, int identity, const char *name, const char *doc, - int unused, const char *signature, + const int unused, const char *signature, PyObject *identity_value) { PyUFuncObject *ufunc; @@ -5220,7 +5200,7 @@ NPY_NO_EXPORT int PyUFunc_RegisterLoopForType(PyUFuncObject *ufunc, int usertype, PyUFuncGenericFunction function, - int *arg_types, + const int *arg_types, void *data) { PyArray_Descr *descr; @@ -5693,18 +5673,13 @@ ufunc_at(PyUFuncObject *ufunc, PyObject *args) * Create dtypes array for either one or two input operands. * The output operand is set to the first input operand */ - dtypes[0] = PyArray_DESCR(op1_array); operands[0] = op1_array; if (op2_array != NULL) { - dtypes[1] = PyArray_DESCR(op2_array); - dtypes[2] = dtypes[0]; operands[1] = op2_array; operands[2] = op1_array; nop = 3; } else { - dtypes[1] = dtypes[0]; - dtypes[2] = NULL; operands[1] = op1_array; operands[2] = NULL; nop = 2; @@ -5861,9 +5836,10 @@ ufunc_at(PyUFuncObject *ufunc, PyObject *args) Py_XDECREF(op2_array); Py_XDECREF(iter); Py_XDECREF(iter2); - Py_XDECREF(array_operands[0]); - Py_XDECREF(array_operands[1]); - Py_XDECREF(array_operands[2]); + for (i = 0; i < 3; i++) { + Py_XDECREF(dtypes[i]); + Py_XDECREF(array_operands[i]); + } if (needs_api && PyErr_Occurred()) { return NULL; @@ -5880,9 +5856,10 @@ fail: Py_XDECREF(op2_array); Py_XDECREF(iter); Py_XDECREF(iter2); - Py_XDECREF(array_operands[0]); - Py_XDECREF(array_operands[1]); - Py_XDECREF(array_operands[2]); + for (i = 0; i < 3; i++) { + Py_XDECREF(dtypes[i]); + Py_XDECREF(array_operands[i]); + } return NULL; } diff --git a/numpy/core/src/umath/ufunc_type_resolution.c b/numpy/core/src/umath/ufunc_type_resolution.c index 25dd002ac..f93d8229e 100644 --- a/numpy/core/src/umath/ufunc_type_resolution.c +++ b/numpy/core/src/umath/ufunc_type_resolution.c @@ -548,6 +548,7 @@ PyUFunc_SimpleUniformOperationTypeResolver( } out_dtypes[0] = ensure_dtype_nbo(dtype); + Py_DECREF(dtype); if (out_dtypes[0] == NULL) { return -1; } @@ -882,7 +883,7 @@ PyUFunc_SubtractionTypeResolver(PyUFuncObject *ufunc, /* The type resolver would have upcast already */ if (out_dtypes[0]->type_num == NPY_BOOL) { PyErr_Format(PyExc_TypeError, - "numpy boolean subtract, the `-` operator, is deprecated, " + "numpy boolean subtract, the `-` operator, is not supported, " "use the bitwise_xor, the `^` operator, or the logical_xor " "function instead."); return -1; @@ -1957,7 +1958,8 @@ linear_search_type_resolver(PyUFuncObject *self, npy_intp i, j, nin = self->nin, nop = nin + self->nout; int types[NPY_MAXARGS]; const char *ufunc_name; - int no_castable_output, use_min_scalar; + int no_castable_output = 0; + int use_min_scalar; /* For making a better error message on coercion error */ char err_dst_typecode = '-', err_src_typecode = '-'; @@ -2264,7 +2266,6 @@ PyUFunc_DivmodTypeResolver(PyUFuncObject *ufunc, out_dtypes[1] = out_dtypes[0]; Py_INCREF(out_dtypes[1]); out_dtypes[2] = PyArray_DescrFromType(NPY_LONGLONG); - Py_INCREF(out_dtypes[2]); out_dtypes[3] = out_dtypes[0]; Py_INCREF(out_dtypes[3]); } diff --git a/numpy/core/tests/test__exceptions.py b/numpy/core/tests/test__exceptions.py new file mode 100644 index 000000000..494b51f34 --- /dev/null +++ b/numpy/core/tests/test__exceptions.py @@ -0,0 +1,42 @@ +""" +Tests of the ._exceptions module. Primarily for exercising the __str__ methods. +""" +import numpy as np + +_ArrayMemoryError = np.core._exceptions._ArrayMemoryError + +class TestArrayMemoryError: + def test_str(self): + e = _ArrayMemoryError((1023,), np.dtype(np.uint8)) + str(e) # not crashing is enough + + # testing these properties is easier than testing the full string repr + def test__size_to_string(self): + """ Test e._size_to_string """ + f = _ArrayMemoryError._size_to_string + Ki = 1024 + assert f(0) == '0 bytes' + assert f(1) == '1 bytes' + assert f(1023) == '1023 bytes' + assert f(Ki) == '1.00 KiB' + assert f(Ki+1) == '1.00 KiB' + assert f(10*Ki) == '10.0 KiB' + assert f(int(999.4*Ki)) == '999. KiB' + assert f(int(1023.4*Ki)) == '1023. KiB' + assert f(int(1023.5*Ki)) == '1.00 MiB' + assert f(Ki*Ki) == '1.00 MiB' + + # 1023.9999 Mib should round to 1 GiB + assert f(int(Ki*Ki*Ki*0.9999)) == '1.00 GiB' + assert f(Ki*Ki*Ki*Ki*Ki*Ki) == '1.00 EiB' + # larger than sys.maxsize, adding larger prefices isn't going to help + # anyway. + assert f(Ki*Ki*Ki*Ki*Ki*Ki*123456) == '123456. EiB' + + def test__total_size(self): + """ Test e._total_size """ + e = _ArrayMemoryError((1,), np.dtype(np.uint8)) + assert e._total_size == 1 + + e = _ArrayMemoryError((2, 4), np.dtype((np.uint64, 16))) + assert e._total_size == 1024 diff --git a/numpy/core/tests/test_api.py b/numpy/core/tests/test_api.py index 32e2ea537..89fc2b0b9 100644 --- a/numpy/core/tests/test_api.py +++ b/numpy/core/tests/test_api.py @@ -296,7 +296,7 @@ def test_array_astype(): ) def test_array_astype_warning(t): # test ComplexWarning when casting from complex to float or int - a = np.array(10, dtype=np.complex) + a = np.array(10, dtype=np.complex_) assert_warns(np.ComplexWarning, a.astype, t) def test_copyto_fromscalar(): diff --git a/numpy/core/tests/test_arrayprint.py b/numpy/core/tests/test_arrayprint.py index f2b8fdca7..702e68e76 100644 --- a/numpy/core/tests/test_arrayprint.py +++ b/numpy/core/tests/test_arrayprint.py @@ -262,11 +262,6 @@ class TestArray2String(object): assert_(np.array2string(s, formatter={'numpystr':lambda s: s*2}) == '[abcabc defdef]') - # check for backcompat that using FloatFormat works and emits warning - with assert_warns(DeprecationWarning): - fmt = np.core.arrayprint.FloatFormat(x, 9, 'maxprec', False) - assert_equal(np.array2string(x, formatter={'float_kind': fmt}), - '[0. 1. 2.]') def test_structure_format(self): dt = np.dtype([('name', np.str_, 16), ('grades', np.float64, (2,))]) @@ -847,9 +842,9 @@ class TestPrintOptions(object): ) def test_bad_args(self): - assert_raises(ValueError, np.set_printoptions, threshold='nan') - assert_raises(ValueError, np.set_printoptions, threshold=u'1') - assert_raises(ValueError, np.set_printoptions, threshold=b'1') + assert_raises(ValueError, np.set_printoptions, threshold=float('nan')) + assert_raises(TypeError, np.set_printoptions, threshold='1') + assert_raises(TypeError, np.set_printoptions, threshold=b'1') def test_unicode_object_array(): import sys diff --git a/numpy/core/tests/test_datetime.py b/numpy/core/tests/test_datetime.py index f99c0f72b..d38444ef7 100644 --- a/numpy/core/tests/test_datetime.py +++ b/numpy/core/tests/test_datetime.py @@ -75,6 +75,15 @@ class TestDateTime(object): # Can cast safely/same_kind from integer to timedelta assert_(np.can_cast('i8', 'm8', casting='same_kind')) assert_(np.can_cast('i8', 'm8', casting='safe')) + assert_(np.can_cast('i4', 'm8', casting='same_kind')) + assert_(np.can_cast('i4', 'm8', casting='safe')) + assert_(np.can_cast('u4', 'm8', casting='same_kind')) + assert_(np.can_cast('u4', 'm8', casting='safe')) + + # Cannot cast safely from unsigned integer of the same size, which + # could overflow + assert_(np.can_cast('u8', 'm8', casting='same_kind')) + assert_(not np.can_cast('u8', 'm8', casting='safe')) # Cannot cast safely/same_kind from float to timedelta assert_(not np.can_cast('f4', 'm8', casting='same_kind')) @@ -136,6 +145,50 @@ class TestDateTime(object): assert_(np.datetime64('NaT') != np.datetime64('NaT', 'us')) assert_(np.datetime64('NaT', 'us') != np.datetime64('NaT')) + @pytest.mark.parametrize("size", [ + 3, 21, 217, 1000]) + def test_datetime_nat_argsort_stability(self, size): + # NaT < NaT should be False internally for + # sort stability + expected = np.arange(size) + arr = np.tile(np.datetime64('NaT'), size) + assert_equal(np.argsort(arr, kind='mergesort'), expected) + + @pytest.mark.parametrize("size", [ + 3, 21, 217, 1000]) + def test_timedelta_nat_argsort_stability(self, size): + # NaT < NaT should be False internally for + # sort stability + expected = np.arange(size) + arr = np.tile(np.timedelta64('NaT'), size) + assert_equal(np.argsort(arr, kind='mergesort'), expected) + + @pytest.mark.parametrize("arr, expected", [ + # the example provided in gh-12629 + (['NaT', 1, 2, 3], + [1, 2, 3, 'NaT']), + # multiple NaTs + (['NaT', 9, 'NaT', -707], + [-707, 9, 'NaT', 'NaT']), + # this sort explores another code path for NaT + ([1, -2, 3, 'NaT'], + [-2, 1, 3, 'NaT']), + # 2-D array + ([[51, -220, 'NaT'], + [-17, 'NaT', -90]], + [[-220, 51, 'NaT'], + [-90, -17, 'NaT']]), + ]) + @pytest.mark.parametrize("dtype", [ + 'M8[ns]', 'M8[us]', + 'm8[ns]', 'm8[us]']) + def test_datetime_timedelta_sort_nat(self, arr, expected, dtype): + # fix for gh-12629 and gh-15063; NaT sorting to end of array + arr = np.array(arr, dtype=dtype) + expected = np.array(expected, dtype=dtype) + arr.sort() + assert_equal(arr, expected) + def test_datetime_scalar_construction(self): # Construct with different units assert_equal(np.datetime64('1950-03-12', 'D'), @@ -483,6 +536,30 @@ class TestDateTime(object): assert_equal(np.datetime64(a, '[Y]'), np.datetime64('NaT', '[Y]')) assert_equal(np.datetime64(a, '[W]'), np.datetime64('NaT', '[W]')) + # NaN -> NaT + nan = np.array([np.nan] * 8) + fnan = nan.astype('f') + lnan = nan.astype('g') + cnan = nan.astype('D') + cfnan = nan.astype('F') + clnan = nan.astype('G') + + nat = np.array([np.datetime64('NaT')] * 8) + assert_equal(nan.astype('M8[ns]'), nat) + assert_equal(fnan.astype('M8[ns]'), nat) + assert_equal(lnan.astype('M8[ns]'), nat) + assert_equal(cnan.astype('M8[ns]'), nat) + assert_equal(cfnan.astype('M8[ns]'), nat) + assert_equal(clnan.astype('M8[ns]'), nat) + + nat = np.array([np.timedelta64('NaT')] * 8) + assert_equal(nan.astype('timedelta64[ns]'), nat) + assert_equal(fnan.astype('timedelta64[ns]'), nat) + assert_equal(lnan.astype('timedelta64[ns]'), nat) + assert_equal(cnan.astype('timedelta64[ns]'), nat) + assert_equal(cfnan.astype('timedelta64[ns]'), nat) + assert_equal(clnan.astype('timedelta64[ns]'), nat) + def test_days_creation(self): assert_equal(np.array('1599', dtype='M8[D]').astype('i8'), (1600-1970)*365 - (1972-1600)/4 + 3 - 365) @@ -1333,10 +1410,14 @@ class TestDateTime(object): # Interaction with NaT a = np.array('1999-03-12T13', dtype='M8[2m]') dtnat = np.array('NaT', dtype='M8[h]') - assert_equal(np.minimum(a, dtnat), a) - assert_equal(np.minimum(dtnat, a), a) - assert_equal(np.maximum(a, dtnat), a) - assert_equal(np.maximum(dtnat, a), a) + assert_equal(np.minimum(a, dtnat), dtnat) + assert_equal(np.minimum(dtnat, a), dtnat) + assert_equal(np.maximum(a, dtnat), dtnat) + assert_equal(np.maximum(dtnat, a), dtnat) + assert_equal(np.fmin(dtnat, a), a) + assert_equal(np.fmin(a, dtnat), a) + assert_equal(np.fmax(dtnat, a), a) + assert_equal(np.fmax(a, dtnat), a) # Also do timedelta a = np.array(3, dtype='m8[h]') @@ -1831,7 +1912,7 @@ class TestDateTime(object): def test_timedelta_arange_no_dtype(self): d = np.array(5, dtype="m8[D]") assert_equal(np.arange(d, d + 1), d) - assert_raises(ValueError, np.arange, d) + assert_equal(np.arange(d), np.arange(0, d)) def test_datetime_maximum_reduce(self): a = np.array(['2010-01-02', '1999-03-14', '1833-03'], dtype='M8[D]') @@ -2208,7 +2289,7 @@ class TestDateTime(object): continue assert_raises(TypeError, np.isnat, np.zeros(10, t)) - def test_isfinite(self): + def test_isfinite_scalar(self): assert_(not np.isfinite(np.datetime64('NaT', 'ms'))) assert_(not np.isfinite(np.datetime64('NaT', 'ns'))) assert_(np.isfinite(np.datetime64('2038-01-19T03:14:07'))) @@ -2216,18 +2297,25 @@ class TestDateTime(object): assert_(not np.isfinite(np.timedelta64('NaT', "ms"))) assert_(np.isfinite(np.timedelta64(34, "ms"))) - res = np.array([True, True, False]) - for unit in ['Y', 'M', 'W', 'D', - 'h', 'm', 's', 'ms', 'us', - 'ns', 'ps', 'fs', 'as']: - arr = np.array([123, -321, "NaT"], dtype='<datetime64[%s]' % unit) - assert_equal(np.isfinite(arr), res) - arr = np.array([123, -321, "NaT"], dtype='>datetime64[%s]' % unit) - assert_equal(np.isfinite(arr), res) - arr = np.array([123, -321, "NaT"], dtype='<timedelta64[%s]' % unit) - assert_equal(np.isfinite(arr), res) - arr = np.array([123, -321, "NaT"], dtype='>timedelta64[%s]' % unit) - assert_equal(np.isfinite(arr), res) + @pytest.mark.parametrize('unit', ['Y', 'M', 'W', 'D', 'h', 'm', 's', 'ms', + 'us', 'ns', 'ps', 'fs', 'as']) + @pytest.mark.parametrize('dstr', ['<datetime64[%s]', '>datetime64[%s]', + '<timedelta64[%s]', '>timedelta64[%s]']) + def test_isfinite_isinf_isnan_units(self, unit, dstr): + '''check isfinite, isinf, isnan for all units of <M, >M, <m, >m dtypes + ''' + arr_val = [123, -321, "NaT"] + arr = np.array(arr_val, dtype= dstr % unit) + pos = np.array([True, True, False]) + neg = np.array([False, False, True]) + false = np.array([False, False, False]) + assert_equal(np.isfinite(arr), pos) + assert_equal(np.isinf(arr), false) + assert_equal(np.isnan(arr), neg) + + def test_assert_equal(self): + assert_raises(AssertionError, assert_equal, + np.datetime64('nat'), np.timedelta64('nat')) def test_corecursive_input(self): # construct a co-recursive list diff --git a/numpy/core/tests/test_deprecations.py b/numpy/core/tests/test_deprecations.py index 58ebea024..363ff26db 100644 --- a/numpy/core/tests/test_deprecations.py +++ b/numpy/core/tests/test_deprecations.py @@ -10,12 +10,16 @@ import sys import operator import warnings import pytest +import shutil +import tempfile import numpy as np from numpy.testing import ( - assert_raises, assert_warns, assert_ + assert_raises, assert_warns, assert_, assert_array_equal ) +from numpy.core._multiarray_tests import fromstring_null_term_c_api + try: import pytz _has_pytz = True @@ -149,16 +153,6 @@ class TestNonTupleNDIndexDeprecation(object): a[[0, 1]] -class TestRankDeprecation(_DeprecationTestCase): - """Test that np.rank is deprecated. The function should simply be - removed. The VisibleDeprecationWarning may become unnecessary. - """ - - def test(self): - a = np.arange(10) - assert_warns(np.VisibleDeprecationWarning, np.rank, a) - - class TestComparisonDeprecations(_DeprecationTestCase): """This tests the deprecation, for non-element-wise comparison logic. This used to mean that when an error occurred during element-wise comparison @@ -178,7 +172,7 @@ class TestComparisonDeprecations(_DeprecationTestCase): # (warning is issued a couple of times here) self.assert_deprecated(op, args=(a, a[:-1]), num=None) - # Element comparison error (numpy array can't be compared). + # ragged array comparison returns True/False a = np.array([1, np.array([1,2,3])], dtype=object) b = np.array([1, np.array([1,2,3])], dtype=object) self.assert_deprecated(op, args=(a, b), num=None) @@ -281,36 +275,6 @@ class TestNonCContiguousViewDeprecation(_DeprecationTestCase): self.assert_deprecated(np.ones((2,2)).T.view, args=(np.int8,)) -class TestInvalidOrderParameterInputForFlattenArrayDeprecation(_DeprecationTestCase): - """Invalid arguments to the ORDER parameter in array.flatten() should not be - allowed and should raise an error. However, in the interests of not breaking - code that may inadvertently pass invalid arguments to this parameter, a - DeprecationWarning will be issued instead for the time being to give developers - time to refactor relevant code. - """ - - def test_flatten_array_non_string_arg(self): - x = np.zeros((3, 5)) - self.message = ("Non-string object detected for " - "the array ordering. Please pass " - "in 'C', 'F', 'A', or 'K' instead") - self.assert_deprecated(x.flatten, args=(np.pi,)) - - def test_flatten_array_invalid_string_arg(self): - # Tests that a DeprecationWarning is raised - # when a string of length greater than one - # starting with "C", "F", "A", or "K" (case- - # and unicode-insensitive) is passed in for - # the ORDER parameter. Otherwise, a TypeError - # will be raised! - - x = np.zeros((3, 5)) - self.message = ("Non length-one string passed " - "in for the array ordering. Please " - "pass in 'C', 'F', 'A', or 'K' instead") - self.assert_deprecated(x.flatten, args=("FACK",)) - - class TestArrayDataAttributeAssignmentDeprecation(_DeprecationTestCase): """Assigning the 'data' attribute of an ndarray is unsafe as pointed out in gh-7093. Eventually, such assignment should NOT be allowed, but @@ -329,22 +293,6 @@ class TestArrayDataAttributeAssignmentDeprecation(_DeprecationTestCase): self.assert_deprecated(a.__setattr__, args=('data', b.data)) -class TestLinspaceInvalidNumParameter(_DeprecationTestCase): - """Argument to the num parameter in linspace that cannot be - safely interpreted as an integer is deprecated in 1.12.0. - - Argument to the num parameter in linspace that cannot be - safely interpreted as an integer should not be allowed. - In the interest of not breaking code that passes - an argument that could still be interpreted as an integer, a - DeprecationWarning will be issued for the time being to give - developers time to refactor relevant code. - """ - def test_float_arg(self): - # 2016-02-25, PR#7328 - self.assert_deprecated(np.linspace, args=(0, 10, 2.5)) - - class TestBinaryReprInsufficientWidthParameterForRepresentation(_DeprecationTestCase): """ If a 'width' parameter is passed into ``binary_repr`` that is insufficient to @@ -452,6 +400,18 @@ class TestNPY_CHAR(_DeprecationTestCase): assert_(npy_char_deprecation() == 'S1') +class TestPyArray_AS1D(_DeprecationTestCase): + def test_npy_pyarrayas1d_deprecation(self): + from numpy.core._multiarray_tests import npy_pyarrayas1d_deprecation + assert_raises(NotImplementedError, npy_pyarrayas1d_deprecation) + + +class TestPyArray_AS2D(_DeprecationTestCase): + def test_npy_pyarrayas2d_deprecation(self): + from numpy.core._multiarray_tests import npy_pyarrayas2d_deprecation + assert_raises(NotImplementedError, npy_pyarrayas2d_deprecation) + + class Test_UPDATEIFCOPY(_DeprecationTestCase): """ v1.14 deprecates creating an array with the UPDATEIFCOPY flag, use @@ -499,6 +459,12 @@ class TestBincount(_DeprecationTestCase): self.assert_deprecated(lambda: np.bincount([1, 2, 3], minlength=None)) +class TestAlen(_DeprecationTestCase): + # 2019-08-02, 1.18.0 + def test_alen(self): + self.assert_deprecated(lambda: np.alen(np.array([1, 2, 3]))) + + class TestGeneratorSum(_DeprecationTestCase): # 2018-02-25, 1.15.0 def test_generator_sum(self): @@ -518,17 +484,71 @@ class TestPositiveOnNonNumerical(_DeprecationTestCase): def test_positive_on_non_number(self): self.assert_deprecated(operator.pos, args=(np.array('foo'),)) + class TestFromstring(_DeprecationTestCase): # 2017-10-19, 1.14 def test_fromstring(self): self.assert_deprecated(np.fromstring, args=('\x00'*80,)) + +class TestFromStringAndFileInvalidData(_DeprecationTestCase): + # 2019-06-08, 1.17.0 + # Tests should be moved to real tests when deprecation is done. + message = "string or file could not be read to its end" + + @pytest.mark.parametrize("invalid_str", [",invalid_data", "invalid_sep"]) + def test_deprecate_unparsable_data_file(self, invalid_str): + x = np.array([1.51, 2, 3.51, 4], dtype=float) + + with tempfile.TemporaryFile(mode="w") as f: + x.tofile(f, sep=',', format='%.2f') + f.write(invalid_str) + + f.seek(0) + self.assert_deprecated(lambda: np.fromfile(f, sep=",")) + f.seek(0) + self.assert_deprecated(lambda: np.fromfile(f, sep=",", count=5)) + # Should not raise: + with warnings.catch_warnings(): + warnings.simplefilter("error", DeprecationWarning) + f.seek(0) + res = np.fromfile(f, sep=",", count=4) + assert_array_equal(res, x) + + @pytest.mark.parametrize("invalid_str", [",invalid_data", "invalid_sep"]) + def test_deprecate_unparsable_string(self, invalid_str): + x = np.array([1.51, 2, 3.51, 4], dtype=float) + x_str = "1.51,2,3.51,4{}".format(invalid_str) + + self.assert_deprecated(lambda: np.fromstring(x_str, sep=",")) + self.assert_deprecated(lambda: np.fromstring(x_str, sep=",", count=5)) + + # The C-level API can use not fixed size, but 0 terminated strings, + # so test that as well: + bytestr = x_str.encode("ascii") + self.assert_deprecated(lambda: fromstring_null_term_c_api(bytestr)) + + with assert_warns(DeprecationWarning): + # this is slightly strange, in that fromstring leaves data + # potentially uninitialized (would be good to error when all is + # read, but count is larger then actual data maybe). + res = np.fromstring(x_str, sep=",", count=5) + assert_array_equal(res[:-1], x) + + with warnings.catch_warnings(): + warnings.simplefilter("error", DeprecationWarning) + + # Should not raise: + res = np.fromstring(x_str, sep=",", count=4) + assert_array_equal(res, x) + + class Test_GetSet_NumericOps(_DeprecationTestCase): # 2018-09-20, 1.16.0 def test_get_numeric_ops(self): from numpy.core._multiarray_tests import getset_numericops self.assert_deprecated(getset_numericops, num=2) - + # empty kwargs prevents any state actually changing which would break # other tests. self.assert_deprecated(np.set_numeric_ops, kwargs={}) diff --git a/numpy/core/tests/test_dtype.py b/numpy/core/tests/test_dtype.py index f60eab696..e18e66c64 100644 --- a/numpy/core/tests/test_dtype.py +++ b/numpy/core/tests/test_dtype.py @@ -25,7 +25,7 @@ def assert_dtype_not_equal(a, b): class TestBuiltin(object): @pytest.mark.parametrize('t', [int, float, complex, np.int32, str, object, - np.unicode]) + np.compat.unicode]) def test_run(self, t): """Only test hash runs at all.""" dt = np.dtype(t) @@ -419,6 +419,31 @@ class TestRecord(object): assert_raises(ValueError, np.dtype, {'formats': ['i4', 'i4'], 'f0': ('i4', 0), 'f1':('i4', 4)}) + def test_fieldless_views(self): + a = np.zeros(2, dtype={'names':[], 'formats':[], 'offsets':[], + 'itemsize':8}) + assert_raises(ValueError, a.view, np.dtype([])) + + d = np.dtype((np.dtype([]), 10)) + assert_equal(d.shape, (10,)) + assert_equal(d.itemsize, 0) + assert_equal(d.base, np.dtype([])) + + arr = np.fromiter((() for i in range(10)), []) + assert_equal(arr.dtype, np.dtype([])) + assert_raises(ValueError, np.frombuffer, b'', dtype=[]) + assert_equal(np.frombuffer(b'', dtype=[], count=2), + np.empty(2, dtype=[])) + + assert_raises(ValueError, np.dtype, ([], 'f8')) + assert_raises(ValueError, np.zeros(1, dtype='i4').view, []) + + assert_equal(np.zeros(2, dtype=[]) == np.zeros(2, dtype=[]), + np.ones(2, dtype=bool)) + + assert_equal(np.zeros((1, 2), dtype=[]) == a, + np.ones((1, 2), dtype=bool)) + class TestSubarray(object): def test_single_subarray(self): @@ -938,13 +963,6 @@ class TestDtypeAttributes(object): new_dtype = np.dtype(dtype.descr) assert_equal(new_dtype.itemsize, 16) - @pytest.mark.parametrize('t', np.typeDict.values()) - def test_name_builtin(self, t): - name = t.__name__ - if name.endswith('_'): - name = name[:-1] - assert_equal(np.dtype(t).name, name) - def test_name_dtype_subclass(self): # Ticket #4357 class user_def_subcls(np.void): @@ -968,7 +986,7 @@ class TestPickling(object): assert_equal(x[0], y[0]) @pytest.mark.parametrize('t', [int, float, complex, np.int32, str, object, - np.unicode, bool]) + np.compat.unicode, bool]) def test_builtin(self, t): self.check_pickling(np.dtype(t)) diff --git a/numpy/core/tests/test_function_base.py b/numpy/core/tests/test_function_base.py index 8b820bd75..c8a7cb6ce 100644 --- a/numpy/core/tests/test_function_base.py +++ b/numpy/core/tests/test_function_base.py @@ -49,7 +49,7 @@ class TestLogspace(object): assert_(len(y) == 50) y = logspace(0, 6, num=100) assert_(y[-1] == 10 ** 6) - y = logspace(0, 6, endpoint=0) + y = logspace(0, 6, endpoint=False) assert_(y[-1] < 10 ** 6) y = logspace(0, 6, num=7) assert_array_equal(y, [1, 10, 100, 1e3, 1e4, 1e5, 1e6]) @@ -229,17 +229,14 @@ class TestLinspace(object): assert_(len(y) == 50) y = linspace(2, 10, num=100) assert_(y[-1] == 10) - y = linspace(2, 10, endpoint=0) + y = linspace(2, 10, endpoint=False) assert_(y[-1] < 10) assert_raises(ValueError, linspace, 0, 10, num=-1) def test_corner(self): y = list(linspace(0, 1, 1)) assert_(y == [0.0], y) - with suppress_warnings() as sup: - sup.filter(DeprecationWarning, ".*safely interpreted as an integer") - y = list(linspace(0, 1, 2.5)) - assert_(y == [0.0, 1.0]) + assert_raises(TypeError, linspace, 0, 1, num=2.5) def test_type(self): t1 = linspace(0, 1, 0).dtype @@ -354,14 +351,20 @@ class TestLinspace(object): arange(j+1, dtype=int)) def test_retstep(self): - y = linspace(0, 1, 2, retstep=True) - assert_(isinstance(y, tuple) and len(y) == 2) - for num in (0, 1): - for ept in (False, True): + for num in [0, 1, 2]: + for ept in [False, True]: y = linspace(0, 1, num, endpoint=ept, retstep=True) - assert_(isinstance(y, tuple) and len(y) == 2 and - len(y[0]) == num and isnan(y[1]), - 'num={0}, endpoint={1}'.format(num, ept)) + assert isinstance(y, tuple) and len(y) == 2 + if num == 2: + y0_expect = [0.0, 1.0] if ept else [0.0, 0.5] + assert_array_equal(y[0], y0_expect) + assert_equal(y[1], y0_expect[1]) + elif num == 1 and not ept: + assert_array_equal(y[0], [0.0]) + assert_equal(y[1], 1.0) + else: + assert_array_equal(y[0], [0.0][:num]) + assert isnan(y[1]) def test_object(self): start = array(1, dtype='O') diff --git a/numpy/core/tests/test_issue14735.py b/numpy/core/tests/test_issue14735.py new file mode 100644 index 000000000..6105c8e6a --- /dev/null +++ b/numpy/core/tests/test_issue14735.py @@ -0,0 +1,29 @@ +import pytest +import warnings +import numpy as np + + +class Wrapper: + def __init__(self, array): + self.array = array + + def __len__(self): + return len(self.array) + + def __getitem__(self, item): + return type(self)(self.array[item]) + + def __getattr__(self, name): + if name.startswith("__array_"): + warnings.warn("object got converted", UserWarning, stacklevel=1) + + return getattr(self.array, name) + + def __repr__(self): + return "<Wrapper({self.array})>".format(self=self) + +@pytest.mark.filterwarnings("error") +def test_getattr_warning(): + array = Wrapper(np.arange(10)) + with pytest.raises(UserWarning, match="object got converted"): + np.asarray(array) diff --git a/numpy/core/tests/test_longdouble.py b/numpy/core/tests/test_longdouble.py index ee4197f8f..2b6e1c5a2 100644 --- a/numpy/core/tests/test_longdouble.py +++ b/numpy/core/tests/test_longdouble.py @@ -5,7 +5,8 @@ import pytest import numpy as np from numpy.testing import ( - assert_, assert_equal, assert_raises, assert_array_equal, temppath, + assert_, assert_equal, assert_raises, assert_warns, assert_array_equal, + temppath, ) from numpy.core.tests._locales import CommaDecimalPointLocale @@ -70,19 +71,54 @@ def test_fromstring(): err_msg="reading '%s'" % s) +def test_fromstring_complex(): + for ctype in ["complex", "cdouble", "cfloat"]: + # Check spacing between separator + assert_equal(np.fromstring("1, 2 , 3 ,4", sep=",", dtype=ctype), + np.array([1., 2., 3., 4.])) + # Real component not specified + assert_equal(np.fromstring("1j, -2j, 3j, 4e1j", sep=",", dtype=ctype), + np.array([1.j, -2.j, 3.j, 40.j])) + # Both components specified + assert_equal(np.fromstring("1+1j,2-2j, -3+3j, -4e1+4j", sep=",", dtype=ctype), + np.array([1. + 1.j, 2. - 2.j, - 3. + 3.j, - 40. + 4j])) + # Spaces at wrong places + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1+2 j,3", dtype=ctype, sep=","), + np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1+ 2j,3", dtype=ctype, sep=","), + np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1 +2j,3", dtype=ctype, sep=","), + np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1+j", dtype=ctype, sep=","), + np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1+", dtype=ctype, sep=","), + np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1j+1", dtype=ctype, sep=","), + np.array([1j])) + + def test_fromstring_bogus(): - assert_equal(np.fromstring("1. 2. 3. flop 4.", dtype=float, sep=" "), - np.array([1., 2., 3.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1. 2. 3. flop 4.", dtype=float, sep=" "), + np.array([1., 2., 3.])) def test_fromstring_empty(): - assert_equal(np.fromstring("xxxxx", sep="x"), - np.array([])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("xxxxx", sep="x"), + np.array([])) def test_fromstring_missing(): - assert_equal(np.fromstring("1xx3x4x5x6", sep="x"), - np.array([1])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1xx3x4x5x6", sep="x"), + np.array([1])) class TestFileBased(object): @@ -95,9 +131,93 @@ class TestFileBased(object): with temppath() as path: with open(path, 'wt') as f: f.write("1. 2. 3. flop 4.\n") - res = np.fromfile(path, dtype=float, sep=" ") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=float, sep=" ") assert_equal(res, np.array([1., 2., 3.])) + def test_fromfile_complex(self): + for ctype in ["complex", "cdouble", "cfloat"]: + # Check spacing between separator and only real component specified + with temppath() as path: + with open(path, 'wt') as f: + f.write("1, 2 , 3 ,4\n") + + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1., 2., 3., 4.])) + + # Real component not specified + with temppath() as path: + with open(path, 'wt') as f: + f.write("1j, -2j, 3j, 4e1j\n") + + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.j, -2.j, 3.j, 40.j])) + + # Both components specified + with temppath() as path: + with open(path, 'wt') as f: + f.write("1+1j,2-2j, -3+3j, -4e1+4j\n") + + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1. + 1.j, 2. - 2.j, - 3. + 3.j, - 40. + 4j])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1+2 j,3\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1+ 2j,3\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1 +2j,3\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1+j\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1+\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.])) + + # Spaces at wrong places + with temppath() as path: + with open(path, 'wt') as f: + f.write("1j+1\n") + + with assert_warns(DeprecationWarning): + res = np.fromfile(path, dtype=ctype, sep=",") + assert_equal(res, np.array([1.j])) + + + @pytest.mark.skipif(string_to_longdouble_inaccurate, reason="Need strtold_l") def test_fromfile(self): @@ -186,12 +306,14 @@ class TestCommaDecimalPointLocale(CommaDecimalPointLocale): assert_equal(a[0], f) def test_fromstring_best_effort_float(self): - assert_equal(np.fromstring("1,234", dtype=float, sep=" "), - np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1,234", dtype=float, sep=" "), + np.array([1.])) def test_fromstring_best_effort(self): - assert_equal(np.fromstring("1,234", dtype=np.longdouble, sep=" "), - np.array([1.])) + with assert_warns(DeprecationWarning): + assert_equal(np.fromstring("1,234", dtype=np.longdouble, sep=" "), + np.array([1.])) def test_fromstring_foreign(self): s = "1.234" @@ -204,8 +326,10 @@ class TestCommaDecimalPointLocale(CommaDecimalPointLocale): assert_array_equal(a, b) def test_fromstring_foreign_value(self): - b = np.fromstring("1,234", dtype=np.longdouble, sep=" ") - assert_array_equal(b[0], 1) + with assert_warns(DeprecationWarning): + b = np.fromstring("1,234", dtype=np.longdouble, sep=" ") + assert_array_equal(b[0], 1) + @pytest.mark.parametrize("int_val", [ # cases discussed in gh-10723 diff --git a/numpy/core/tests/test_multiarray.py b/numpy/core/tests/test_multiarray.py index 6fa8548a0..22d550ecc 100644 --- a/numpy/core/tests/test_multiarray.py +++ b/numpy/core/tests/test_multiarray.py @@ -20,6 +20,7 @@ import gc import weakref import pytest from contextlib import contextmanager +from test.support import no_tracing from numpy.compat import pickle @@ -114,7 +115,7 @@ class TestFlags(object): # Ensure that any base being writeable is sufficient to change flag; # this is especially interesting for arrays from an array interface. arr = np.arange(10) - + class subclass(np.ndarray): pass @@ -964,7 +965,7 @@ class TestCreation(object): @pytest.mark.skipif(sys.version_info[0] >= 3, reason="Not Python 2") def test_sequence_long(self): - assert_equal(np.array([long(4), long(4)]).dtype, np.long) + assert_equal(np.array([long(4), long(4)]).dtype, long) assert_equal(np.array([long(4), 2**80]).dtype, object) assert_equal(np.array([long(4), 2**80, long(4)]).dtype, object) assert_equal(np.array([2**80, long(4)]).dtype, object) @@ -1784,7 +1785,7 @@ class TestMethods(object): # test unicode sorts. s = 'aaaaaaaa' - a = np.array([s + chr(i) for i in range(101)], dtype=np.unicode) + a = np.array([s + chr(i) for i in range(101)], dtype=np.unicode_) b = a[::-1].copy() for kind in self.sort_kinds: msg = "unicode sort, kind=%s" % kind @@ -2036,7 +2037,7 @@ class TestMethods(object): # test unicode argsorts. s = 'aaaaaaaa' - a = np.array([s + chr(i) for i in range(101)], dtype=np.unicode) + a = np.array([s + chr(i) for i in range(101)], dtype=np.unicode_) b = a[::-1] r = np.arange(101) rr = r[::-1] @@ -2119,7 +2120,7 @@ class TestMethods(object): a = np.array(['aaaaaaaaa' for i in range(100)]) assert_equal(a.argsort(kind='m'), r) # unicode - a = np.array(['aaaaaaaaa' for i in range(100)], dtype=np.unicode) + a = np.array(['aaaaaaaaa' for i in range(100)], dtype=np.unicode_) assert_equal(a.argsort(kind='m'), r) def test_sort_unicode_kind(self): @@ -2248,7 +2249,7 @@ class TestMethods(object): 'P:\\20x_dapi_cy3\\20x_dapi_cy3_20100197_1', 'P:\\20x_dapi_cy3\\20x_dapi_cy3_20100198_1', 'P:\\20x_dapi_cy3\\20x_dapi_cy3_20100199_1'], - dtype=np.unicode) + dtype=np.unicode_) ind = np.arange(len(a)) assert_equal([a.searchsorted(v, 'left') for v in a], ind) assert_equal([a.searchsorted(v, 'right') for v in a], ind + 1) @@ -2766,6 +2767,12 @@ class TestMethods(object): assert_equal(x1.flatten('F'), y1f) assert_equal(x1.flatten('F'), x1.T.flatten()) + def test_flatten_invalid_order(self): + # invalid after gh-14596 + for order in ['Z', 'c', False, True, 0, 8]: + x = np.array([[1, 2, 3], [4, 5, 6]], np.int32) + assert_raises(ValueError, x.flatten, {"order": order}) + @pytest.mark.parametrize('func', (np.dot, np.matmul)) def test_arr_mult(self, func): a = np.array([[1, 0], [0, 1]]) @@ -3573,10 +3580,10 @@ class TestBinop(object): assert_equal(np.modf(dummy, out=(None, a)), (1,)) assert_equal(np.modf(dummy, out=(dummy, a)), (1,)) assert_equal(np.modf(a, out=(dummy, a)), 0) - with warnings.catch_warnings(record=True) as w: - warnings.filterwarnings('always', '', DeprecationWarning) - assert_equal(np.modf(dummy, out=a), (0,)) - assert_(w[0].category is DeprecationWarning) + with assert_raises(TypeError): + # Out argument must be tuple, since there are multiple outputs + np.modf(dummy, out=a) + assert_raises(ValueError, np.modf, dummy, out=(a,)) # 2 inputs, 1 output @@ -3941,13 +3948,13 @@ class TestPickling(object): def test_datetime64_byteorder(self): original = np.array([['2015-02-24T00:00:00.000000000']], dtype='datetime64[ns]') - + original_byte_reversed = original.copy(order='K') original_byte_reversed.dtype = original_byte_reversed.dtype.newbyteorder('S') original_byte_reversed.byteswap(inplace=True) new = pickle.loads(pickle.dumps(original_byte_reversed)) - + assert_equal(original.dtype, new.dtype) @@ -4076,17 +4083,17 @@ class TestArgmax(object): np.datetime64('2010-01-03T05:14:12'), np.datetime64('NaT'), np.datetime64('2015-09-23T10:10:13'), - np.datetime64('1932-10-10T03:50:30')], 4), + np.datetime64('1932-10-10T03:50:30')], 0), ([np.datetime64('2059-03-14T12:43:12'), np.datetime64('1996-09-21T14:43:15'), np.datetime64('NaT'), np.datetime64('2022-12-25T16:02:16'), np.datetime64('1963-10-04T03:14:12'), - np.datetime64('2013-05-08T18:15:23')], 0), + np.datetime64('2013-05-08T18:15:23')], 2), ([np.timedelta64(2, 's'), np.timedelta64(1, 's'), np.timedelta64('NaT', 's'), - np.timedelta64(3, 's')], 3), + np.timedelta64(3, 's')], 2), ([np.timedelta64('NaT', 's')] * 3, 0), ([timedelta(days=5, seconds=14), timedelta(days=2, seconds=35), @@ -4160,6 +4167,7 @@ class TestArgmax(object): assert_equal(a.argmax(out=out1, axis=0), np.argmax(a, out=out2, axis=0)) assert_equal(out1, out2) + @pytest.mark.leaks_references(reason="replaces None with NULL.") def test_object_argmax_with_NULLs(self): # See gh-6032 a = np.empty(4, dtype='O') @@ -4210,17 +4218,17 @@ class TestArgmin(object): np.datetime64('2010-01-03T05:14:12'), np.datetime64('NaT'), np.datetime64('2015-09-23T10:10:13'), - np.datetime64('1932-10-10T03:50:30')], 5), + np.datetime64('1932-10-10T03:50:30')], 0), ([np.datetime64('2059-03-14T12:43:12'), np.datetime64('1996-09-21T14:43:15'), np.datetime64('NaT'), np.datetime64('2022-12-25T16:02:16'), np.datetime64('1963-10-04T03:14:12'), - np.datetime64('2013-05-08T18:15:23')], 4), + np.datetime64('2013-05-08T18:15:23')], 2), ([np.timedelta64(2, 's'), np.timedelta64(1, 's'), np.timedelta64('NaT', 's'), - np.timedelta64(3, 's')], 1), + np.timedelta64(3, 's')], 2), ([np.timedelta64('NaT', 's')] * 3, 0), ([timedelta(days=5, seconds=14), timedelta(days=2, seconds=35), @@ -4308,6 +4316,7 @@ class TestArgmin(object): assert_equal(a.argmin(out=out1, axis=0), np.argmin(a, out=out2, axis=0)) assert_equal(out1, out2) + @pytest.mark.leaks_references(reason="replaces None with NULL.") def test_object_argmin_with_NULLs(self): # See gh-6032 a = np.empty(4, dtype='O') @@ -4335,18 +4344,14 @@ class TestMinMax(object): assert_equal(np.amax([[1, 2, 3]], axis=1), 3) def test_datetime(self): - # NaTs are ignored + # Do not ignore NaT for dtype in ('m8[s]', 'm8[Y]'): a = np.arange(10).astype(dtype) - a[3] = 'NaT' assert_equal(np.amin(a), a[0]) assert_equal(np.amax(a), a[9]) - a[0] = 'NaT' - assert_equal(np.amin(a), a[1]) - assert_equal(np.amax(a), a[9]) - a.fill('NaT') - assert_equal(np.amin(a), a[0]) - assert_equal(np.amax(a), a[0]) + a[3] = 'NaT' + assert_equal(np.amin(a), a[3]) + assert_equal(np.amax(a), a[3]) class TestNewaxis(object): @@ -4848,7 +4853,7 @@ class TestIO(object): offset_bytes = self.dtype.itemsize z = np.fromfile(f, dtype=self.dtype, offset=offset_bytes) assert_array_equal(z, self.x.flat[offset_items+count_items+1:]) - + with open(self.filename, 'wb') as f: self.x.tofile(f, sep=",") @@ -4938,7 +4943,8 @@ class TestIO(object): self._check_from(b'1,2,3,4', [1., 2., 3., 4.], dtype=float, sep=',') def test_malformed(self): - self._check_from(b'1.234 1,234', [1.234, 1.], sep=' ') + with assert_warns(DeprecationWarning): + self._check_from(b'1.234 1,234', [1.234, 1.], sep=' ') def test_long_sep(self): self._check_from(b'1_x_3_x_4_x_5', [1, 3, 4, 5], sep='_x_') @@ -4991,6 +4997,19 @@ class TestIO(object): self.test_tofile_sep() self.test_tofile_format() + def test_fromfile_subarray_binary(self): + # Test subarray dtypes which are absorbed into the shape + x = np.arange(24, dtype="i4").reshape(2, 3, 4) + x.tofile(self.filename) + res = np.fromfile(self.filename, dtype="(3,4)i4") + assert_array_equal(x, res) + + x_str = x.tobytes() + with assert_warns(DeprecationWarning): + # binary fromstring is deprecated + res = np.fromstring(x_str, dtype="(3,4)i4") + assert_array_equal(x, res) + class TestFromBuffer(object): @pytest.mark.parametrize('byteorder', ['<', '>']) @@ -5078,6 +5097,8 @@ class TestFlat(object): class TestResize(object): + + @no_tracing def test_basic(self): x = np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]]) if IS_PYPY: @@ -5094,6 +5115,7 @@ class TestResize(object): assert_raises(ValueError, x.resize, (5, 1)) del y # avoid pyflakes unused variable warning. + @no_tracing def test_int_shape(self): x = np.eye(3) if IS_PYPY: @@ -5127,6 +5149,7 @@ class TestResize(object): assert_raises(TypeError, np.eye(3).resize, order=1) assert_raises(TypeError, np.eye(3).resize, refcheck='hi') + @no_tracing def test_freeform_shape(self): x = np.eye(3) if IS_PYPY: @@ -5135,6 +5158,7 @@ class TestResize(object): x.resize(3, 2, 1) assert_(x.shape == (3, 2, 1)) + @no_tracing def test_zeros_appended(self): x = np.eye(3) if IS_PYPY: @@ -5144,6 +5168,7 @@ class TestResize(object): assert_array_equal(x[0], np.eye(3)) assert_array_equal(x[1], np.zeros((3, 3))) + @no_tracing def test_obj_obj(self): # check memory is initialized on resize, gh-4857 a = np.ones(10, dtype=[('k', object, 2)]) @@ -6205,14 +6230,14 @@ class TestMatmul(MatmulCommon): r3 = np.matmul(args[0].copy(), args[1].copy()) assert_equal(r1, r3) - + def test_matmul_object(self): import fractions f = np.vectorize(fractions.Fraction) def random_ints(): return np.random.randint(1, 1000, size=(10, 3, 3)) - M1 = f(random_ints(), random_ints()) + M1 = f(random_ints(), random_ints()) M2 = f(random_ints(), random_ints()) M3 = self.matmul(M1, M2) @@ -6402,20 +6427,22 @@ class TestInner(object): class TestAlen(object): def test_basic(self): - m = np.array([1, 2, 3]) - assert_equal(np.alen(m), 3) + with pytest.warns(DeprecationWarning): + m = np.array([1, 2, 3]) + assert_equal(np.alen(m), 3) - m = np.array([[1, 2, 3], [4, 5, 7]]) - assert_equal(np.alen(m), 2) + m = np.array([[1, 2, 3], [4, 5, 7]]) + assert_equal(np.alen(m), 2) - m = [1, 2, 3] - assert_equal(np.alen(m), 3) + m = [1, 2, 3] + assert_equal(np.alen(m), 3) - m = [[1, 2, 3], [4, 5, 7]] - assert_equal(np.alen(m), 2) + m = [[1, 2, 3], [4, 5, 7]] + assert_equal(np.alen(m), 2) def test_singleton(self): - assert_equal(np.alen(5), 1) + with pytest.warns(DeprecationWarning): + assert_equal(np.alen(5), 1) class TestChoose(object): @@ -7202,6 +7229,13 @@ class TestNewBufferProtocol(object): RuntimeError, "ndim", np.array, m) + # The above seems to create some deep cycles, clean them up for + # easier reference count debugging: + del c_u8_33d, m + for i in range(33): + if gc.collect() == 0: + break + def test_error_pointer_type(self): # gh-6741 m = memoryview(ctypes.pointer(ctypes.c_uint8())) @@ -7746,6 +7780,7 @@ if not IS_PYPY: d = np.ones(100) assert_(sys.getsizeof(d) < sys.getsizeof(d.reshape(100, 1, 1).copy())) + @no_tracing def test_resize(self): d = np.ones(100) old = sys.getsizeof(d) @@ -7880,20 +7915,20 @@ class TestBytestringArrayNonzero(object): class TestUnicodeArrayNonzero(object): def test_empty_ustring_array_is_falsey(self): - assert_(not np.array([''], dtype=np.unicode)) + assert_(not np.array([''], dtype=np.unicode_)) def test_whitespace_ustring_array_is_falsey(self): - a = np.array(['eggs'], dtype=np.unicode) + a = np.array(['eggs'], dtype=np.unicode_) a[0] = ' \0\0' assert_(not a) def test_all_null_ustring_array_is_falsey(self): - a = np.array(['eggs'], dtype=np.unicode) + a = np.array(['eggs'], dtype=np.unicode_) a[0] = '\0\0\0\0' assert_(not a) def test_null_inside_ustring_array_is_truthy(self): - a = np.array(['eggs'], dtype=np.unicode) + a = np.array(['eggs'], dtype=np.unicode_) a[0] = ' \0 \0' assert_(a) @@ -8094,6 +8129,8 @@ class TestWritebackIfCopy(object): arr_wb[...] = 100 assert_equal(arr, -100) + @pytest.mark.leaks_references( + reason="increments self in dealloc; ignore since deprecated path.") def test_dealloc_warning(self): with suppress_warnings() as sup: sup.record(RuntimeWarning) diff --git a/numpy/core/tests/test_nditer.py b/numpy/core/tests/test_nditer.py index cf66751f8..daec9ce6d 100644 --- a/numpy/core/tests/test_nditer.py +++ b/numpy/core/tests/test_nditer.py @@ -2104,7 +2104,7 @@ def test_iter_buffering_string(): assert_equal(i[0], b'abc') assert_equal(i[0].dtype, np.dtype('S6')) - a = np.array(['abc', 'a', 'abcd'], dtype=np.unicode) + a = np.array(['abc', 'a', 'abcd'], dtype=np.unicode_) assert_equal(a.dtype, np.dtype('U4')) assert_raises(TypeError, nditer, a, ['buffered'], ['readonly'], op_dtypes='U2') diff --git a/numpy/core/tests/test_numeric.py b/numpy/core/tests/test_numeric.py index 935b84234..ffebdf648 100644 --- a/numpy/core/tests/test_numeric.py +++ b/numpy/core/tests/test_numeric.py @@ -1254,6 +1254,39 @@ class TestNonzero(object): a = np.array([[False], [TrueThenFalse()]]) assert_raises(RuntimeError, np.nonzero, a) + def test_nonzero_exception_safe(self): + # gh-13930 + + class ThrowsAfter: + def __init__(self, iters): + self.iters_left = iters + + def __bool__(self): + if self.iters_left == 0: + raise ValueError("called `iters` times") + + self.iters_left -= 1 + return True + + """ + Test that a ValueError is raised instead of a SystemError + + If the __bool__ function is called after the error state is set, + Python (cpython) will raise a SystemError. + """ + + # assert that an exception in first pass is handled correctly + a = np.array([ThrowsAfter(5)]*10) + assert_raises(ValueError, np.nonzero, a) + + # raise exception in second pass for 1-dimensional loop + a = np.array([ThrowsAfter(15)]*10) + assert_raises(ValueError, np.nonzero, a) + + # raise exception in second pass for n-dimensional loop + a = np.array([[ThrowsAfter(15)]]*10) + assert_raises(ValueError, np.nonzero, a) + class TestIndex(object): def test_boolean(self): @@ -1308,6 +1341,11 @@ class TestBinaryRepr(object): exp = '1' + (width - 1) * '0' assert_equal(np.binary_repr(num, width=width), exp) + def test_large_neg_int64(self): + # See gh-14289. + assert_equal(np.binary_repr(np.int64(-2**62), width=64), + '11' + '0'*62) + class TestBaseRepr(object): def test_base3(self): @@ -2529,6 +2567,11 @@ class TestCorrelate(object): z = np.correlate(y, x, mode='full') assert_array_almost_equal(z, r_z) + def test_zero_size(self): + with pytest.raises(ValueError): + np.correlate(np.array([]), np.ones(1000), mode='full') + with pytest.raises(ValueError): + np.correlate(np.ones(1000), np.array([]), mode='full') class TestConvolve(object): def test_object(self): @@ -2545,6 +2588,30 @@ class TestConvolve(object): class TestArgwhere(object): + + @pytest.mark.parametrize('nd', [0, 1, 2]) + def test_nd(self, nd): + # get an nd array with multiple elements in every dimension + x = np.empty((2,)*nd, bool) + + # none + x[...] = False + assert_equal(np.argwhere(x).shape, (0, nd)) + + # only one + x[...] = False + x.flat[0] = True + assert_equal(np.argwhere(x).shape, (1, nd)) + + # all but one + x[...] = True + x.flat[0] = False + assert_equal(np.argwhere(x).shape, (x.size - 1, nd)) + + # all + x[...] = True + assert_equal(np.argwhere(x).shape, (x.size, nd)) + def test_2D(self): x = np.arange(6).reshape((2, 3)) assert_array_equal(np.argwhere(x > 1), @@ -2886,7 +2953,7 @@ class TestIndices(object): assert_array_equal(x, np.array([[0], [1], [2], [3]])) assert_array_equal(y, np.array([[0, 1, 2]])) - @pytest.mark.parametrize("dtype", [np.int, np.float32, np.float64]) + @pytest.mark.parametrize("dtype", [np.int32, np.int64, np.float32, np.float64]) @pytest.mark.parametrize("dims", [(), (0,), (4, 3)]) def test_return_type(self, dtype, dims): inds = np.indices(dims, dtype=dtype) diff --git a/numpy/core/tests/test_numerictypes.py b/numpy/core/tests/test_numerictypes.py index d0ff5578a..387740e35 100644 --- a/numpy/core/tests/test_numerictypes.py +++ b/numpy/core/tests/test_numerictypes.py @@ -498,3 +498,32 @@ class TestDocStrings(object): assert_('int64' in np.int_.__doc__) elif np.int64 is np.longlong: assert_('int64' in np.longlong.__doc__) + + +class TestScalarTypeNames: + # gh-9799 + + numeric_types = [ + np.byte, np.short, np.intc, np.int_, np.longlong, + np.ubyte, np.ushort, np.uintc, np.uint, np.ulonglong, + np.half, np.single, np.double, np.longdouble, + np.csingle, np.cdouble, np.clongdouble, + ] + + def test_names_are_unique(self): + # none of the above may be aliases for each other + assert len(set(self.numeric_types)) == len(self.numeric_types) + + # names must be unique + names = [t.__name__ for t in self.numeric_types] + assert len(set(names)) == len(names) + + @pytest.mark.parametrize('t', numeric_types) + def test_names_reflect_attributes(self, t): + """ Test that names correspond to where the type is under ``np.`` """ + assert getattr(np, t.__name__) is t + + @pytest.mark.parametrize('t', numeric_types) + def test_names_are_undersood_by_dtype(self, t): + """ Test the dtype constructor maps names back to the type """ + assert np.dtype(t.__name__).type is t diff --git a/numpy/core/tests/test_regression.py b/numpy/core/tests/test_regression.py index cbd4ceafc..3880b1394 100644 --- a/numpy/core/tests/test_regression.py +++ b/numpy/core/tests/test_regression.py @@ -17,6 +17,7 @@ from numpy.testing import ( _assert_valid_refcount, HAS_REFCOUNT, ) from numpy.compat import asbytes, asunicode, long, pickle +from test.support import no_tracing try: RecursionError @@ -1316,6 +1317,7 @@ class TestRegression(object): assert_(pickle.loads( pickle.dumps(test_record, protocol=proto)) == test_record) + @no_tracing def test_blasdot_uninitialized_memory(self): # Ticket #950 for m in [0, 1, 2]: @@ -1511,7 +1513,7 @@ class TestRegression(object): min //= -1 with np.errstate(divide="ignore"): - for t in (np.int8, np.int16, np.int32, np.int64, int, np.long): + for t in (np.int8, np.int16, np.int32, np.int64, int, np.compat.long): test_type(t) def test_buffer_hashlib(self): @@ -1539,7 +1541,8 @@ class TestRegression(object): def test_fromstring_crash(self): # Ticket #1345: the following should not cause a crash - np.fromstring(b'aa, aa, 1.0', sep=',') + with assert_warns(DeprecationWarning): + np.fromstring(b'aa, aa, 1.0', sep=',') def test_ticket_1539(self): dtypes = [x for x in np.typeDict.values() @@ -2111,7 +2114,7 @@ class TestRegression(object): # Ticket #1578, the mismatch only showed up when running # python-debug for python versions >= 2.7, and then as # a core dump and error message. - a = np.array(['abc'], dtype=np.unicode)[0] + a = np.array(['abc'], dtype=np.unicode_)[0] del a def test_refcount_error_in_clip(self): @@ -2481,26 +2484,6 @@ class TestRegression(object): np.array([T()]) - def test_2d__array__shape(self): - class T(object): - def __array__(self): - return np.ndarray(shape=(0,0)) - - # Make sure __array__ is used instead of Sequence methods. - def __iter__(self): - return iter([]) - - def __getitem__(self, idx): - raise AssertionError("__getitem__ was called") - - def __len__(self): - return 0 - - - t = T() - #gh-13659, would raise in broadcasting [x=t for x in result] - np.array([t]) - @pytest.mark.skipif(sys.maxsize < 2 ** 31 + 1, reason='overflows 32-bit python') @pytest.mark.skipif(sys.platform == 'win32' and sys.version_info[:2] < (3, 8), reason='overflows on windows, fixed in bpo-16865') diff --git a/numpy/core/tests/test_scalarinherit.py b/numpy/core/tests/test_scalarinherit.py index 9e32cf624..6a5c4fde9 100644 --- a/numpy/core/tests/test_scalarinherit.py +++ b/numpy/core/tests/test_scalarinherit.py @@ -68,8 +68,7 @@ class TestCharacter(object): def test_char_repeat(self): np_s = np.string_('abc') np_u = np.unicode_('abc') - np_i = np.int(5) res_s = b'abc' * 5 res_u = u'abc' * 5 - assert_(np_s * np_i == res_s) - assert_(np_u * np_i == res_u) + assert_(np_s * 5 == res_s) + assert_(np_u * 5 == res_u) diff --git a/numpy/core/tests/test_scalarmath.py b/numpy/core/tests/test_scalarmath.py index ebba457e3..c84380cd9 100644 --- a/numpy/core/tests/test_scalarmath.py +++ b/numpy/core/tests/test_scalarmath.py @@ -11,7 +11,7 @@ import numpy as np from numpy.testing import ( assert_, assert_equal, assert_raises, assert_almost_equal, assert_array_equal, IS_PYPY, suppress_warnings, _gen_alignment_data, - assert_warns + assert_warns, assert_raises_regex, ) types = [np.bool_, np.byte, np.ubyte, np.short, np.ushort, np.intc, np.uintc, @@ -293,6 +293,16 @@ class TestModulus(object): rem = operator.mod(finf, fone) assert_(np.isnan(rem), 'dt: %s' % dt) + def test_inplace_floordiv_handling(self): + # issue gh-12927 + # this only applies to in-place floordiv //=, because the output type + # promotes to float which does not fit + a = np.array([1, 2], np.int64) + b = np.array([1, 2], np.uint64) + pattern = 'could not be coerced to provided output parameter' + with assert_raises_regex(TypeError, pattern): + a //= b + class TestComplexDivision(object): def test_zero_division(self): @@ -664,3 +674,31 @@ class TestAbs(object): def test_numpy_abs(self): self._test_abs_func(np.abs) + + +class TestBitShifts(object): + + @pytest.mark.parametrize('type_code', np.typecodes['AllInteger']) + @pytest.mark.parametrize('op', + [operator.rshift, operator.lshift], ids=['>>', '<<']) + def test_shift_all_bits(self, type_code, op): + """ Shifts where the shift amount is the width of the type or wider """ + # gh-2449 + dt = np.dtype(type_code) + nbits = dt.itemsize * 8 + for val in [5, -5]: + for shift in [nbits, nbits + 4]: + val_scl = dt.type(val) + shift_scl = dt.type(shift) + res_scl = op(val_scl, shift_scl) + if val_scl < 0 and op is operator.rshift: + # sign bit is preserved + assert_equal(res_scl, -1) + else: + assert_equal(res_scl, 0) + + # Result on scalars should be the same as on arrays + val_arr = np.array([val]*32, dtype=dt) + shift_arr = np.array([shift]*32, dtype=dt) + res_arr = op(val_arr, shift_arr) + assert_equal(res_arr, res_scl) diff --git a/numpy/core/tests/test_ufunc.py b/numpy/core/tests/test_ufunc.py index 707c690dd..526925ece 100644 --- a/numpy/core/tests/test_ufunc.py +++ b/numpy/core/tests/test_ufunc.py @@ -980,7 +980,7 @@ class TestUfunc(object): assert_array_equal(out, mm_row_col_vec.squeeze()) def test_matrix_multiply(self): - self.compare_matrix_multiply_results(np.long) + self.compare_matrix_multiply_results(np.int64) self.compare_matrix_multiply_results(np.double) def test_matrix_multiply_umath_empty(self): @@ -1092,7 +1092,6 @@ class TestUfunc(object): arr0d = np.array(HasComparisons()) assert_equal(arr0d == arr0d, True) assert_equal(np.equal(arr0d, arr0d), True) # normal behavior is a cast - assert_equal(np.equal(arr0d, arr0d, dtype=object), '==') arr1d = np.array([HasComparisons()]) assert_equal(arr1d == arr1d, np.array([True])) diff --git a/numpy/core/tests/test_umath.py b/numpy/core/tests/test_umath.py index 1d93f9ac0..e892e81d2 100644 --- a/numpy/core/tests/test_umath.py +++ b/numpy/core/tests/test_umath.py @@ -75,11 +75,9 @@ class TestOut(object): assert_(r1 is o1) assert_(r2 is o2) - with warnings.catch_warnings(record=True) as w: - warnings.filterwarnings('always', '', DeprecationWarning) + with assert_raises(TypeError): + # Out argument must be tuple, since there are multiple outputs. r1, r2 = np.frexp(d, out=o1, subok=subok) - assert_(r1 is o1) - assert_(w[0].category is DeprecationWarning) assert_raises(ValueError, np.add, a, 2, o, o, subok=subok) assert_raises(ValueError, np.add, a, 2, o, out=o, subok=subok) @@ -165,19 +163,14 @@ class TestOut(object): else: assert_(type(r1) == np.ndarray) - with warnings.catch_warnings(record=True) as w: - warnings.filterwarnings('always', '', DeprecationWarning) + with assert_raises(TypeError): + # Out argument must be tuple, since there are multiple outputs. r1, r2 = np.frexp(d, out=o1, subok=subok) - if subok: - assert_(isinstance(r2, ArrayWrap)) - else: - assert_(type(r2) == np.ndarray) - assert_(w[0].category is DeprecationWarning) class TestComparisons(object): def test_ignore_object_identity_in_equal(self): - # Check error raised when comparing identical objects whose comparison + # Check comparing identical objects whose comparison # is not a simple boolean, e.g., arrays that are compared elementwise. a = np.array([np.array([1, 2, 3]), None], dtype=object) assert_raises(ValueError, np.equal, a, a) @@ -195,7 +188,7 @@ class TestComparisons(object): assert_equal(np.equal(a, a), [False]) def test_ignore_object_identity_in_not_equal(self): - # Check error raised when comparing identical objects whose comparison + # Check comparing identical objects whose comparison # is not a simple boolean, e.g., arrays that are compared elementwise. a = np.array([np.array([1, 2, 3]), None], dtype=object) assert_raises(ValueError, np.not_equal, a, a) @@ -678,30 +671,151 @@ class TestSpecialFloats(object): assert_raises(FloatingPointError, np.log, np.float32(-np.inf)) assert_raises(FloatingPointError, np.log, np.float32(-1.0)) -class TestExpLogFloat32(object): + def test_sincos_values(self): + with np.errstate(all='ignore'): + x = [np.nan, np.nan, np.nan, np.nan] + y = [np.nan, -np.nan, np.inf, -np.inf] + for dt in ['f', 'd', 'g']: + xf = np.array(x, dtype=dt) + yf = np.array(y, dtype=dt) + assert_equal(np.sin(yf), xf) + assert_equal(np.cos(yf), xf) + + with np.errstate(invalid='raise'): + assert_raises(FloatingPointError, np.sin, np.float32(-np.inf)) + assert_raises(FloatingPointError, np.sin, np.float32(np.inf)) + assert_raises(FloatingPointError, np.cos, np.float32(-np.inf)) + assert_raises(FloatingPointError, np.cos, np.float32(np.inf)) + + def test_sqrt_values(self): + with np.errstate(all='ignore'): + x = [np.nan, np.nan, np.inf, np.nan, 0.] + y = [np.nan, -np.nan, np.inf, -np.inf, 0.] + for dt in ['f', 'd', 'g']: + xf = np.array(x, dtype=dt) + yf = np.array(y, dtype=dt) + assert_equal(np.sqrt(yf), xf) + + #with np.errstate(invalid='raise'): + # for dt in ['f', 'd', 'g']: + # assert_raises(FloatingPointError, np.sqrt, np.array(-100., dtype=dt)) + + def test_abs_values(self): + x = [np.nan, np.nan, np.inf, np.inf, 0., 0., 1.0, 1.0] + y = [np.nan, -np.nan, np.inf, -np.inf, 0., -0., -1.0, 1.0] + for dt in ['f', 'd', 'g']: + xf = np.array(x, dtype=dt) + yf = np.array(y, dtype=dt) + assert_equal(np.abs(yf), xf) + + def test_square_values(self): + x = [np.nan, np.nan, np.inf, np.inf] + y = [np.nan, -np.nan, np.inf, -np.inf] + with np.errstate(all='ignore'): + for dt in ['f', 'd', 'g']: + xf = np.array(x, dtype=dt) + yf = np.array(y, dtype=dt) + assert_equal(np.square(yf), xf) + + with np.errstate(over='raise'): + assert_raises(FloatingPointError, np.square, np.array(1E32, dtype='f')) + assert_raises(FloatingPointError, np.square, np.array(1E200, dtype='d')) + + def test_reciprocal_values(self): + with np.errstate(all='ignore'): + x = [np.nan, np.nan, 0.0, -0.0, np.inf, -np.inf] + y = [np.nan, -np.nan, np.inf, -np.inf, 0., -0.] + for dt in ['f', 'd', 'g']: + xf = np.array(x, dtype=dt) + yf = np.array(y, dtype=dt) + assert_equal(np.reciprocal(yf), xf) + + with np.errstate(divide='raise'): + for dt in ['f', 'd', 'g']: + assert_raises(FloatingPointError, np.reciprocal, np.array(-0.0, dtype=dt)) + +# func : [maxulperror, low, high] +avx_ufuncs = {'sqrt' :[1, 0., 100.], + 'absolute' :[0, -100., 100.], + 'reciprocal' :[1, 1., 100.], + 'square' :[1, -100., 100.], + 'rint' :[0, -100., 100.], + 'floor' :[0, -100., 100.], + 'ceil' :[0, -100., 100.], + 'trunc' :[0, -100., 100.]} + +class TestAVXUfuncs(object): + def test_avx_based_ufunc(self): + strides = np.array([-4,-3,-2,-1,1,2,3,4]) + np.random.seed(42) + for func, prop in avx_ufuncs.items(): + maxulperr = prop[0] + minval = prop[1] + maxval = prop[2] + # various array sizes to ensure masking in AVX is tested + for size in range(1,32): + myfunc = getattr(np, func) + x_f32 = np.float32(np.random.uniform(low=minval, high=maxval, + size=size)) + x_f64 = np.float64(x_f32) + x_f128 = np.longdouble(x_f32) + y_true128 = myfunc(x_f128) + if maxulperr == 0: + assert_equal(myfunc(x_f32), np.float32(y_true128)) + assert_equal(myfunc(x_f64), np.float64(y_true128)) + else: + assert_array_max_ulp(myfunc(x_f32), np.float32(y_true128), + maxulp=maxulperr) + assert_array_max_ulp(myfunc(x_f64), np.float64(y_true128), + maxulp=maxulperr) + # various strides to test gather instruction + if size > 1: + y_true32 = myfunc(x_f32) + y_true64 = myfunc(x_f64) + for jj in strides: + assert_equal(myfunc(x_f64[::jj]), y_true64[::jj]) + assert_equal(myfunc(x_f32[::jj]), y_true32[::jj]) + +class TestAVXFloat32Transcendental(object): def test_exp_float32(self): np.random.seed(42) x_f32 = np.float32(np.random.uniform(low=0.0,high=88.1,size=1000000)) x_f64 = np.float64(x_f32) - assert_array_max_ulp(np.exp(x_f32), np.float32(np.exp(x_f64)), maxulp=2.6) + assert_array_max_ulp(np.exp(x_f32), np.float32(np.exp(x_f64)), maxulp=3) def test_log_float32(self): np.random.seed(42) x_f32 = np.float32(np.random.uniform(low=0.0,high=1000,size=1000000)) x_f64 = np.float64(x_f32) - assert_array_max_ulp(np.log(x_f32), np.float32(np.log(x_f64)), maxulp=3.9) + assert_array_max_ulp(np.log(x_f32), np.float32(np.log(x_f64)), maxulp=4) + + def test_sincos_float32(self): + np.random.seed(42) + N = 1000000 + M = np.int_(N/20) + index = np.random.randint(low=0, high=N, size=M) + x_f32 = np.float32(np.random.uniform(low=-100.,high=100.,size=N)) + # test coverage for elements > 117435.992f for which glibc is used + x_f32[index] = np.float32(10E+10*np.random.rand(M)) + x_f64 = np.float64(x_f32) + assert_array_max_ulp(np.sin(x_f32), np.float32(np.sin(x_f64)), maxulp=2) + assert_array_max_ulp(np.cos(x_f32), np.float32(np.cos(x_f64)), maxulp=2) - def test_strided_exp_log_float32(self): + def test_strided_float32(self): np.random.seed(42) - strides = np.random.randint(low=-100, high=100, size=100) - sizes = np.random.randint(low=1, high=2000, size=100) + strides = np.array([-4,-3,-2,-1,1,2,3,4]) + sizes = np.arange(2,100) for ii in sizes: x_f32 = np.float32(np.random.uniform(low=0.01,high=88.1,size=ii)) exp_true = np.exp(x_f32) log_true = np.log(x_f32) + sin_true = np.sin(x_f32) + cos_true = np.cos(x_f32) for jj in strides: assert_array_almost_equal_nulp(np.exp(x_f32[::jj]), exp_true[::jj], nulp=2) assert_array_almost_equal_nulp(np.log(x_f32[::jj]), log_true[::jj], nulp=2) + assert_array_almost_equal_nulp(np.sin(x_f32[::jj]), sin_true[::jj], nulp=2) + assert_array_almost_equal_nulp(np.cos(x_f32[::jj]), cos_true[::jj], nulp=2) class TestLogAddExp(_FilterInvalids): def test_logaddexp_values(self): @@ -1655,7 +1769,6 @@ class TestSpecialMethods(object): ok = np.empty(1).view(Ok) bad = np.empty(1).view(Bad) - # double-free (segfault) of "ok" if "bad" raises an exception for i in range(10): assert_raises(RuntimeError, ncu.frexp, 1, ok, bad) @@ -2129,10 +2242,9 @@ class TestSpecialMethods(object): assert_(np.modf(a, None) == {}) assert_(np.modf(a, None, None) == {}) assert_(np.modf(a, out=(None, None)) == {}) - with warnings.catch_warnings(record=True) as w: - warnings.filterwarnings('always', '', DeprecationWarning) - assert_(np.modf(a, out=None) == {}) - assert_(w[0].category is DeprecationWarning) + with assert_raises(TypeError): + # Out argument must be tuple, since there are multiple outputs. + np.modf(a, out=None) # don't give positional and output argument, or too many arguments. # wrong number of arguments in the tuple is an error too. diff --git a/numpy/core/tests/test_umath_accuracy.py b/numpy/core/tests/test_umath_accuracy.py index fcbed0dd3..fec180786 100644 --- a/numpy/core/tests/test_umath_accuracy.py +++ b/numpy/core/tests/test_umath_accuracy.py @@ -35,9 +35,10 @@ class TestAccuracy(object): for filename in files: data_dir = path.join(path.dirname(__file__), 'data') filepath = path.join(data_dir, filename) - file_without_comments = (r for r in open(filepath) if not r[0] in ('$', '#')) + with open(filepath) as fid: + file_without_comments = (r for r in fid if not r[0] in ('$', '#')) data = np.genfromtxt(file_without_comments, - dtype=('|S39','|S39','|S39',np.int), + dtype=('|S39','|S39','|S39',int), names=('type','input','output','ulperr'), delimiter=',', skip_header=1) |