1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
|
from pypy.interpreter.error import oefmt
from rpython.rlib import jit
from pypy.module.micronumpy import constants as NPY
from pypy.module.micronumpy.base import W_NDimArray
# structures to describe slicing
class BaseChunk(object):
_attrs_ = ['step', 'out_dim']
class Chunk(BaseChunk):
input_dim = 1
def __init__(self, start, stop, step, lgt):
self.start = start
self.stop = stop
self.step = step
self.lgt = lgt
if self.step == 0:
self.out_dim = 0
else:
self.out_dim = 1
def compute(self, space, base_length, base_stride):
stride = base_stride * self.step
backstride = base_stride * max(0, self.lgt - 1) * self.step
return self.start, self.lgt, stride, backstride
def __repr__(self):
return 'Chunk(%d, %d, %d, %d)' % (self.start, self.stop, self.step,
self.lgt)
class IntegerChunk(BaseChunk):
input_dim = 1
out_dim = 0
def __init__(self, w_idx):
self.w_idx = w_idx
def compute(self, space, base_length, base_stride):
start, _, _, _ = space.decode_index4_unsafe(self.w_idx, base_length)
return start, 0, 0, 0
class SliceChunk(BaseChunk):
input_dim = 1
out_dim = 1
def __init__(self, w_slice):
self.w_slice = w_slice
def compute(self, space, base_length, base_stride):
start, stop, step, length = space.decode_index4_unsafe(self.w_slice, base_length)
stride = base_stride * step
backstride = base_stride * max(0, length - 1) * step
return start, length, stride, backstride
class NewAxisChunk(Chunk):
input_dim = 0
out_dim = 1
def __init__(self):
pass
def compute(self, space, base_length, base_stride):
return 0, 1, 0, 0
class EllipsisChunk(BaseChunk):
input_dim = 0
out_dim = 0
def __init__(self):
pass
def compute(self, space, base_length, base_stride):
backstride = base_stride * max(0, base_length - 1)
return 0, base_length, base_stride, backstride
class BooleanChunk(BaseChunk):
input_dim = 1
out_dim = 1
def __init__(self, w_idx):
self.w_idx = w_idx
def compute(self, space, base_length, base_stride):
raise oefmt(space.w_NotImplementedError, 'cannot reach')
def new_view(space, w_arr, chunks):
arr = w_arr.implementation
dim = -1
for i, c in enumerate(chunks):
if isinstance(c, BooleanChunk):
dim = i
break
if dim >= 0:
# filter by axis dim
filtr = chunks[dim]
assert isinstance(filtr, BooleanChunk)
# XXX this creates a new array, and fails in setitem
w_arr = w_arr.getitem_filter(space, filtr.w_idx, axis=dim)
arr = w_arr.implementation
chunks[dim] = SliceChunk(space.newslice(space.newint(0),
space.w_None, space.w_None))
r = calculate_slice_strides(space, arr.shape, arr.start,
arr.get_strides(), arr.get_backstrides(), chunks)
else:
r = calculate_slice_strides(space, arr.shape, arr.start,
arr.get_strides(), arr.get_backstrides(), chunks)
shape, start, strides, backstrides = r
return W_NDimArray.new_slice(space, start, strides[:], backstrides[:],
shape[:], arr, w_arr)
@jit.unroll_safe
def _extend_shape(old_shape, chunks):
shape = []
i = -1
for i, c in enumerate_chunks(chunks):
if c.out_dim > 0:
shape.append(c.lgt)
s = i + 1
assert s >= 0
return shape[:] + old_shape[s:]
class BaseTransform(object):
pass
class ViewTransform(BaseTransform):
def __init__(self, chunks):
# 4-tuple specifying slicing
self.chunks = chunks
class BroadcastTransform(BaseTransform):
def __init__(self, res_shape):
self.res_shape = res_shape
@jit.look_inside_iff(lambda chunks: jit.isconstant(len(chunks)))
def enumerate_chunks(chunks):
result = []
i = -1
for chunk in chunks:
i += chunk.input_dim
result.append((i, chunk))
return result
@jit.look_inside_iff(lambda space, shape, start, strides, backstrides, chunks:
jit.isconstant(len(chunks)))
def calculate_slice_strides(space, shape, start, strides, backstrides, chunks):
"""
Note: `chunks` can contain at most one EllipsisChunk object.
"""
size = 0
used_dims = 0
for chunk in chunks:
used_dims += chunk.input_dim
size += chunk.out_dim
if used_dims > len(shape):
raise oefmt(space.w_IndexError, "too many indices for array")
else:
extra_dims = len(shape) - used_dims
rstrides = [0] * (size + extra_dims)
rbackstrides = [0] * (size + extra_dims)
rshape = [0] * (size + extra_dims)
rstart = start
i = 0 # index of the current dimension in the input array
j = 0 # index of the current dimension in the result view
for chunk in chunks:
if isinstance(chunk, NewAxisChunk):
rshape[j] = 1
j += 1
continue
elif isinstance(chunk, EllipsisChunk):
for k in range(extra_dims):
start, length, stride, backstride = chunk.compute(
space, shape[i], strides[i])
rshape[j] = length
rstrides[j] = stride
rbackstrides[j] = backstride
j += 1
i += 1
continue
start, length, stride, backstride = chunk.compute(space, shape[i], strides[i])
if chunk.out_dim == 1:
rshape[j] = length
rstrides[j] = stride
rbackstrides[j] = backstride
j += chunk.out_dim
rstart += strides[i] * start
i += chunk.input_dim
return rshape, rstart, rstrides, rbackstrides
def calculate_broadcast_strides(strides, backstrides, orig_shape, res_shape, backwards=False):
rstrides = []
rbackstrides = []
for i in range(len(orig_shape)):
if orig_shape[i] == 1:
rstrides.append(0)
rbackstrides.append(0)
else:
rstrides.append(strides[i])
rbackstrides.append(backstrides[i])
if backwards:
rstrides = rstrides + [0] * (len(res_shape) - len(orig_shape))
rbackstrides = rbackstrides + [0] * (len(res_shape) - len(orig_shape))
else:
rstrides = [0] * (len(res_shape) - len(orig_shape)) + rstrides
rbackstrides = [0] * (len(res_shape) - len(orig_shape)) + rbackstrides
return rstrides, rbackstrides
@jit.unroll_safe
def shape_agreement(space, shape1, w_arr2, broadcast_down=True):
if w_arr2 is None:
return shape1
assert isinstance(w_arr2, W_NDimArray)
shape2 = w_arr2.get_shape()
ret = _shape_agreement(shape1, shape2)
if len(ret) < max(len(shape1), len(shape2)):
def format_shape(shape):
if len(shape) > 1:
return ",".join([str(x) for x in shape])
else:
return '%d,' % shape[0]
raise oefmt(space.w_ValueError,
"operands could not be broadcast together with shapes "
"(%s) (%s)", format_shape(shape1), format_shape(shape2))
if not broadcast_down and len([x for x in ret if x != 1]) > len([x for x in shape2 if x != 1]):
raise oefmt(space.w_ValueError,
"unbroadcastable shape (%s) cannot be broadcasted to (%s)",
",".join([str(x) for x in shape1]),
",".join([str(x) for x in shape2])
)
return ret
@jit.unroll_safe
def shape_agreement_multiple(space, array_list, shape=None):
""" call shape_agreement recursively, allow elements from array_list to
be None (like w_out)
"""
for arr in array_list:
if not space.is_none(arr):
if shape is None:
shape = arr.get_shape()
else:
shape = shape_agreement(space, shape, arr)
return shape
@jit.unroll_safe
def _shape_agreement(shape1, shape2):
""" Checks agreement about two shapes with respect to broadcasting. Returns
the resulting shape.
"""
lshift = 0
rshift = 0
if len(shape1) > len(shape2):
m = len(shape1)
n = len(shape2)
rshift = len(shape2) - len(shape1)
remainder = shape1
else:
m = len(shape2)
n = len(shape1)
lshift = len(shape1) - len(shape2)
remainder = shape2
endshape = [0] * m
indices1 = [True] * m
indices2 = [True] * m
for i in range(m - 1, m - n - 1, -1):
left = shape1[i + lshift]
right = shape2[i + rshift]
if left == right:
endshape[i] = left
elif left == 1:
endshape[i] = right
indices1[i + lshift] = False
elif right == 1:
endshape[i] = left
indices2[i + rshift] = False
else:
return []
#raise oefmt(space.w_ValueError,
# "frames are not aligned")
for i in range(m - n):
endshape[i] = remainder[i]
return endshape
def get_shape_from_iterable(space, old_size, w_iterable):
new_size = 0
new_shape = []
if space.isinstance_w(w_iterable, space.w_int):
new_size = space.int_w(w_iterable)
if new_size < 0:
new_size = old_size
new_shape = [new_size]
else:
neg_dim = -1
batch = space.listview(w_iterable)
new_size = 1
new_shape = []
i = 0
for elem in batch:
s = space.int_w(elem)
if s < 0:
if neg_dim >= 0:
raise oefmt(space.w_ValueError,
"can only specify one unknown dimension")
s = 1
neg_dim = i
new_size *= s
new_shape.append(s)
i += 1
if neg_dim >= 0:
new_shape[neg_dim] = old_size / new_size
new_size *= new_shape[neg_dim]
if new_size != old_size:
raise oefmt(space.w_ValueError,
"total size of new array must be unchanged")
return new_shape
@jit.unroll_safe
def calc_strides(shape, dtype, order):
strides = []
backstrides = []
s = 1
shape_rev = shape[:]
if order in [NPY.CORDER, NPY.ANYORDER]:
shape_rev.reverse()
for sh in shape_rev:
slimit = max(sh, 1)
strides.append(s * dtype.elsize)
backstrides.append(s * (slimit - 1) * dtype.elsize)
s *= slimit
if order in [NPY.CORDER, NPY.ANYORDER]:
strides.reverse()
backstrides.reverse()
return strides, backstrides
@jit.unroll_safe
def calc_backstrides(strides, shape):
ndims = len(shape)
new_backstrides = [0] * ndims
for nd in range(ndims):
new_backstrides[nd] = (shape[nd] - 1) * strides[nd]
return new_backstrides
# Recalculating strides. Find the steps that the iteration does for each
# dimension, given the stride and shape. Then try to create a new stride that
# fits the new shape, using those steps. If there is a shape/step mismatch
# (meaning that the realignment of elements crosses from one step into another)
# return None so that the caller can raise an exception.
def calc_new_strides(new_shape, old_shape, old_strides, order):
# Return the proper strides for new_shape, or None if the mapping crosses
# stepping boundaries
# Assumes that prod(old_shape) == prod(new_shape), len(old_shape) > 1, and
# len(new_shape) > 0
steps = []
last_step = 1
oldI = 0
new_strides = []
if order == NPY.FORTRANORDER:
for i in range(len(old_shape)):
steps.append(old_strides[i] / last_step)
last_step *= old_shape[i]
cur_step = steps[0]
n_new_elems_used = 1
n_old_elems_to_use = old_shape[0]
for s in new_shape:
new_strides.append(cur_step * n_new_elems_used)
n_new_elems_used *= s
while n_new_elems_used > n_old_elems_to_use:
oldI += 1
if steps[oldI] != steps[oldI - 1]:
return None
n_old_elems_to_use *= old_shape[oldI]
if n_new_elems_used == n_old_elems_to_use:
oldI += 1
if oldI < len(old_shape):
cur_step = steps[oldI]
n_old_elems_to_use *= old_shape[oldI]
else:
for i in range(len(old_shape) - 1, -1, -1):
steps.insert(0, old_strides[i] / last_step)
last_step *= old_shape[i]
cur_step = steps[-1]
n_new_elems_used = 1
oldI = -1
n_old_elems_to_use = old_shape[-1]
for i in range(len(new_shape) - 1, -1, -1):
s = new_shape[i]
new_strides.insert(0, cur_step * n_new_elems_used)
n_new_elems_used *= s
while n_new_elems_used > n_old_elems_to_use:
oldI -= 1
if steps[oldI] != steps[oldI + 1]:
return None
n_old_elems_to_use *= old_shape[oldI]
if n_new_elems_used == n_old_elems_to_use:
oldI -= 1
if oldI >= -len(old_shape):
cur_step = steps[oldI]
n_old_elems_to_use *= old_shape[oldI]
return new_strides[:]
def calc_start(shape, strides):
''' Strides can be negative for non-contiguous data.
Calculate the appropriate positive starting position so
the indexing still works properly
'''
start = 0
for i in range(len(shape)):
if strides[i] < 0:
start -= strides[i] * (shape[i] - 1)
return start
@jit.unroll_safe
def is_c_contiguous(arr):
shape = arr.get_shape()
strides = arr.get_strides()
ret = True
sd = arr.dtype.elsize
for i in range(len(shape) - 1, -1, -1):
dim = shape[i]
if strides[i] != sd:
ret = False
break
if dim == 0:
break
sd *= dim
return ret
@jit.unroll_safe
def is_f_contiguous(arr):
shape = arr.get_shape()
strides = arr.get_strides()
ret = True
sd = arr.dtype.elsize
for i in range(len(shape)):
dim = shape[i]
if strides[i] != sd:
ret = False
break
if dim == 0:
break
sd *= dim
return ret
|