def _correlate_kernel(Array, Filter, mode):
# Anything to do?
if Array.size <= 0:
return Array.copy()
# Complex correlate includes a conjugation
if numpy.iscomplexobj(Filter):
Filter = numpy.conj(Filter)
# Get sizes as arrays for easier manipulation
ArrSize = numpy.array(Array.shape, dtype=numpy.int32)
FilterSize = numpy.array(Filter.shape, dtype=numpy.int32)
# Check that mode='valid' is allowed given the array sizes
if mode == 'valid':
diffSize = ArrSize[:FilterSize.size] - FilterSize
nSmaller = summations.sum(diffSize < 0)
if nSmaller > 0:
raise ValueError(
"correlateNd: For 'valid' mode, one must be at least as large as the other in every dimension")
# Use numpy convention for result dype
dtype = numpy.result_type(Array, Filter)
# Add zeros along relevant dimensions
Padded = _addZerosNd(Array, FilterSize, dtype)
PaddedSize = numpy.array(Padded.shape, dtype=numpy.int32)
# Get positions of first view
IndiVec = _findIndices(PaddedSize, FilterSize)
CenterPos = tuple((FilterSize - 1) // 2)
IndiMat = IndiVec.reshape(FilterSize, order='F')
nPre = IndiMat[CenterPos] # Required zeros before Array for correct alignment
nPost = IndiVec[Filter.size - 1] - nPre
n = Padded.size
nTot = n + nPre + nPost # Total size after pre/post padding
V = array_create.empty([nTot], dtype=dtype, bohrium=bhary.check(Array))
V[nPre:n + nPre] = Padded.flatten(order='F')
V[:nPre] = 0
V[n + nPre:] = 0
A = Filter.flatten(order='F')
# Actual correlation calculation
Correlated = V[IndiVec[0]:n + IndiVec[0]] * A[0]
for i in range(1, Filter.size):
Correlated += V[IndiVec[i]:n + IndiVec[i]] * A[i]
# TODO: we need this flush because of very slow fusion
if bhary.check(V):
_bh.flush()
Full = Correlated.reshape(PaddedSize, order='F')
if mode == 'full':
return Full
elif mode == 'same':
return _findSame(Full, FilterSize)
elif mode == 'valid':
return _findValid(Full, FilterSize)
else:
raise ValueError("correlateNd: invalid mode '%s'" % mode)
def _addZerosNd(Array, FilterSize, dtype):
# Introduces zero padding for Column major flattening
PaddedSize = numpy.array(Array.shape, dtype=numpy.int32)
N = FilterSize.shape[0]
PaddedSize[0:N] += FilterSize - 1
cut = '['
for i in range(PaddedSize.shape[0]):
if i < N:
minpos = int(FilterSize[i] / 2)
maxpos = Array.shape[i] + int(FilterSize[i] / 2)
else:
minpos = 0
maxpos = Array.shape[i]
cut += str(minpos) + ':' + str(maxpos) + ','
cut = cut[:-1] + ']'
Padded = array_create.zeros(PaddedSize, dtype=dtype, bohrium=bhary.check(Array))
exec ('Padded' + cut + '=Array')
return Padded
def _addZerosNd(Array, FilterSize, dtype):
# Introduces zero padding for Column major flattening
PaddedSize = numpy.array(Array.shape, dtype=numpy.int32)
N = FilterSize.shape[0]
PaddedSize[0:N] += FilterSize - 1
cut = '['
for i in range(PaddedSize.shape[0]):
if i < N:
minpos = int(FilterSize[i] / 2)
maxpos = Array.shape[i] + int(FilterSize[i] / 2)
else:
minpos = 0
maxpos = Array.shape[i]
cut += str(minpos) + ':' + str(maxpos) + ','
cut = cut[:-1] + ']'
Padded = array_create.zeros(PaddedSize,
dtype=dtype,
bohrium=bhary.check(Array))
exec('Padded' + cut + '=Array')
return Padded
def array(obj, dtype=None, copy=False, order=None, subok=False, ndmin=0, bohrium=True):
"""
Create an array -- Bohrium or NumPy ndarray.
Parameters
----------
obj : array_like
An array, any object exposing the array interface, an
object whose __array__ method returns an array, or any
(nested) sequence.
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.
copy : bool, optional
If true, then the object is copied. Otherwise, a copy
will only be made if __array__ returns a copy, if obj is a
nested sequence, or if a copy is needed to satisfy any of the other
requirements (`dtype`, `order`, etc.).
order : {'C', 'F', 'A'}, optional
Specify the order of the array. If order is 'C' (default), then the
array will be in C-contiguous order (last-index varies the
fastest). If order is 'F', then the returned array
will be in Fortran-contiguous order (first-index varies the
fastest). If order is 'A', then the returned array may
be in any order (either C-, Fortran-contiguous, or even
discontiguous).
subok : bool, optional
If True, then sub-classes will be passed-through, otherwise
the returned array will be forced to be a base-class array (default).
ndmin : int, optional
Specifies the minimum number of dimensions that the resulting
array should have. Ones will be pre-pended to the shape as
needed to meet this requirement.
bohrium : boolean, optional
Determines whether it is a Bohrium array (bohrium.ndarray) or a
regular NumPy array (numpy.ndarray)
Returns
-------
out : ndarray
An array object satisfying the specified requirements.
See Also
--------
empty, empty_like, zeros, zeros_like, ones, ones_like, fill
Examples
--------
>>> np.array([1, 2, 3])
array([1, 2, 3])
Upcasting:
>>> np.array([1, 2, 3.0])
array([ 1., 2., 3.])
More than one dimension:
>>> np.array([[1, 2], [3, 4]])
array([[1, 2],
[3, 4]])
Minimum dimensions 2:
>>> np.array([1, 2, 3], ndmin=2)
array([[1, 2, 3]])
Type provided:
>>> np.array([1, 2, 3], dtype=complex)
array([ 1.+0.j, 2.+0.j, 3.+0.j])
Data-type consisting of more than one element:
>>> x = np.array([(1,2),(3,4)],dtype=[('a','<i4'),('b','<i4')])
>>> x['a']
array([1, 3])
Creating an array from sub-classes:
>>> np.array(np.mat('1 2; 3 4'))
array([[1, 2],
[3, 4]])
>>> np.array(np.mat('1 2; 3 4'), subok=True)
matrix([[1, 2],
[3, 4]])
"""
ary = obj
if bohrium:
if bhary.check(ary):
if order == 'F':
raise ValueError("Cannot convert a Bohrium array to column-major ('F') memory representation")
elif order == 'C' and not ary.flags['C_CONTIGUOUS']:
copy = True # We need to copy in order to make the returned array contiguous
if copy:
t = empty_like(ary)
t[...] = ary
ary = t
if dtype is not None and not dtype_equal(dtype, ary.dtype):
t = empty_like(ary, dtype=dtype)
t[...] = ary
ary = t
for i in range(ary.ndim, ndmin):
ary = numpy.expand_dims(ary, i)
return ary
else:
# Let's convert the array using regular NumPy.
# When `ary` is not a regular NumPy array, we make sure that `ary` contains no Bohrium arrays
if isinstance(ary, collections.Sequence) and \
not (isinstance(ary, numpy.ndarray) and ary.dtype.isbuiltin == 1):
ary = list(ary) # Let's make sure that `ary` is mutable
for i in range(len(ary)): # Converting 1-element Bohrium arrays to NumPy scalars
if bhary.check(ary[i]):
ary[i] = ary[i].copy2numpy()
ary = numpy.array(ary, dtype=dtype, copy=copy, order=order, subok=subok, ndmin=ndmin, fix_biclass=False)
# In any case, the array must meet some requirements
ary = numpy.require(ary, requirements=['C_CONTIGUOUS', 'ALIGNED', 'OWNDATA'])
if bohrium and not dtype_support(ary.dtype):
_warn_dtype(ary.dtype, 3)
return ary
ret = empty(ary.shape, dtype=ary.dtype)
if ret.size > 0:
ret._data_fill(ary)
return ret
else:
if bhary.check(ary):
ret = ary.copy2numpy()
return numpy.array(ret, dtype=dtype, copy=copy, order=order, subok=subok, ndmin=ndmin, fix_biclass=False)
else:
return numpy.array(ary, dtype=dtype, copy=copy, order=order, subok=subok, ndmin=ndmin, fix_biclass=False)
def _correlate_kernel(Array, Filter, mode):
# Anything to do?
if Array.size <= 0:
return Array.copy()
# Complex correlate includes a conjugation
if numpy.iscomplexobj(Filter):
Filter = numpy.conj(Filter)
# Get sizes as arrays for easier manipulation
ArrSize = numpy.array(Array.shape, dtype=numpy.int32)
FilterSize = numpy.array(Filter.shape, dtype=numpy.int32)
# Check that mode='valid' is allowed given the array sizes
if mode == 'valid':
diffSize = ArrSize[:FilterSize.size] - FilterSize
nSmaller = summations.sum(diffSize < 0)
if nSmaller > 0:
raise ValueError(
"correlateNd: For 'valid' mode, one must be at least as large as the other in every dimension"
)
# Use numpy convention for result dype
dtype = numpy.result_type(Array, Filter)
# Add zeros along relevant dimensions
Padded = _addZerosNd(Array, FilterSize, dtype)
PaddedSize = numpy.array(Padded.shape, dtype=numpy.int32)
# Get positions of first view
IndiVec = _findIndices(PaddedSize, FilterSize)
CenterPos = tuple((FilterSize - 1) // 2)
IndiMat = IndiVec.reshape(FilterSize, order='F')
nPre = IndiMat[
CenterPos] # Required zeros before Array for correct alignment
nPost = IndiVec[Filter.size - 1] - nPre
n = Padded.size
nTot = n + nPre + nPost # Total size after pre/post padding
V = array_create.empty([nTot], dtype=dtype, bohrium=bhary.check(Array))
V[nPre:n + nPre] = Padded.flatten(order='F')
V[:nPre] = 0
V[n + nPre:] = 0
A = Filter.flatten(order='F')
# Actual correlation calculation
Correlated = V[IndiVec[0]:n + IndiVec[0]] * A[0]
for i in range(1, Filter.size):
Correlated += V[IndiVec[i]:n + IndiVec[i]] * A[i]
# TODO: we need this flush because of very slow fusion
if bhary.check(V):
_util.flush()
Full = Correlated.reshape(PaddedSize, order='F')
if mode == 'full':
return Full
elif mode == 'same':
return _findSame(Full, FilterSize)
elif mode == 'valid':
return _findValid(Full, FilterSize)
else:
raise ValueError("correlateNd: invalid mode '%s'" % mode)
def as_strided(x, shape=None, strides=None, subok=True, writeable=True):
"""
Create a view into the array with the given shape and strides.
.. warning:: This function has to be used with extreme care, see notes.
Parameters
----------
x : ndarray
Array to create a new.
shape : sequence of int, optional
The shape of the new array. Defaults to ``x.shape``.
strides : sequence of int, optional
The strides of the new array. Defaults to ``x.strides``.
subok : bool, optional
.. versionadded:: 1.10
If True, subclasses are preserved.
writeable : bool, optional
.. versionadded:: 1.12
If set to False, the returned array will always be readonly.
Otherwise it will be writable if the original array was. It
is advisable to set this to False if possible (see Notes).
Returns
-------
view : ndarray
See also
--------
broadcast_to: broadcast an array to a given shape.
reshape : reshape an array.
Notes
-----
``as_strided`` creates a view into the array given the exact strides
and shape. This means it manipulates the internal data structure of
ndarray and, if done incorrectly, the array elements can point to
invalid memory and can corrupt results or crash your program.
It is advisable to always use the original ``x.strides`` when
calculating new strides to avoid reliance on a contiguous memory
layout.
Furthermore, arrays created with this function often contain self
overlapping memory, so that two elements are identical.
Vectorized write operations on such arrays will typically be
unpredictable. They may even give different results for small, large,
or transposed arrays.
Since writing to these arrays has to be tested and done with great
care, you may want to use ``writeable=False`` to avoid accidental write
operations.
For these reasons it is advisable to avoid ``as_strided`` when
possible.
"""
class DummyArray(object):
"""Dummy object that just exists to hang __array_interface__ dictionaries
and possibly keep alive a reference to a base array.
"""
def __init__(self, interface, base=None):
self.__array_interface__ = interface
self.base = base
def _maybe_view_as_subclass(original_array, new_array):
if type(original_array) is not type(new_array):
# if input was an ndarray subclass and subclasses were OK,
# then view the result as that subclass.
new_array = new_array.view(type=type(original_array))
# Since we have done something akin to a view from original_array, we
# should let the subclass finalize (if it has it implemented, i.e., is
# not None).
if new_array.__array_finalize__:
new_array.__array_finalize__(original_array)
return new_array
# first convert input to array, possibly keeping subclass
x = numpy.array(x, copy=False, subok=subok)
interface = dict(x.__array_interface__)
if shape is not None:
interface['shape'] = tuple(shape)
if strides is not None:
interface['strides'] = tuple(strides)
array = numpy.asarray(DummyArray(interface, base=x))
# The route via `__interface__` does not preserve structured
# dtypes. Since dtype should remain unchanged, we set it explicitly.
array.dtype = x.dtype
view = _maybe_view_as_subclass(x, array)
if view.flags.writeable and not writeable:
view.flags.writeable = False
return view
def as_strided(x, shape=None, strides=None, subok=True, writeable=True):
"""
Create a view into the array with the given shape and strides.
.. warning:: This function has to be used with extreme care, see notes.
Parameters
----------
x : ndarray
Array to create a new.
shape : sequence of int, optional
The shape of the new array. Defaults to ``x.shape``.
strides : sequence of int, optional
The strides of the new array. Defaults to ``x.strides``.
subok : bool, optional
.. versionadded:: 1.10
If True, subclasses are preserved.
writeable : bool, optional
.. versionadded:: 1.12
If set to False, the returned array will always be readonly.
Otherwise it will be writable if the original array was. It
is advisable to set this to False if possible (see Notes).
Returns
-------
view : ndarray
See also
--------
broadcast_to: broadcast an array to a given shape.
reshape : reshape an array.
Notes
-----
``as_strided`` creates a view into the array given the exact strides
and shape. This means it manipulates the internal data structure of
ndarray and, if done incorrectly, the array elements can point to
invalid memory and can corrupt results or crash your program.
It is advisable to always use the original ``x.strides`` when
calculating new strides to avoid reliance on a contiguous memory
layout.
Furthermore, arrays created with this function often contain self
overlapping memory, so that two elements are identical.
Vectorized write operations on such arrays will typically be
unpredictable. They may even give different results for small, large,
or transposed arrays.
Since writing to these arrays has to be tested and done with great
care, you may want to use ``writeable=False`` to avoid accidental write
operations.
For these reasons it is advisable to avoid ``as_strided`` when
possible.
"""
class DummyArray(object):
"""Dummy object that just exists to hang __array_interface__ dictionaries
and possibly keep alive a reference to a base array.
"""
def __init__(self, interface, base=None):
self.__array_interface__ = interface
self.base = base
def _maybe_view_as_subclass(original_array, new_array):
if type(original_array) is not type(new_array):
# if input was an ndarray subclass and subclasses were OK,
# then view the result as that subclass.
new_array = new_array.view(type=type(original_array))
# Since we have done something akin to a view from original_array, we
# should let the subclass finalize (if it has it implemented, i.e., is
# not None).
if new_array.__array_finalize__:
new_array.__array_finalize__(original_array)
return new_array
# first convert input to array, possibly keeping subclass
x = numpy.array(x, copy=False, subok=subok)
interface = dict(x.__array_interface__)
if shape is not None:
interface['shape'] = tuple(shape)
if strides is not None:
interface['strides'] = tuple(strides)
array = numpy.asarray(DummyArray(interface, base=x))
# The route via `__interface__` does not preserve structured
# dtypes. Since dtype should remain unchanged, we set it explicitly.
array.dtype = x.dtype
view = _maybe_view_as_subclass(x, array)
if view.flags.writeable and not writeable:
view.flags.writeable = False
return view