Skip to content

P
pytensor
提交 33c4b2cc authored 作者: Frederic's avatar Frederic
浏览文件
操作

parametrise the helper fct in Subtensor to allow reusing it in GpuSubtensor.

Also add missing decref in error handling code.
上级 ae13b006
隐藏空白字符变更
内嵌 并排
正在显示 包含 61 行增加 和 24 行删除
theano/tensor/basic.py
浏览文件 @ 33c4b2cc
......@@ -3952,9 +3952,21 @@ class Subtensor(Op):
return "%s{%s}" % (self.__class__.__name__, ", ".join(indices))
@staticmethod
def helper_c_code(node, name, inputs, outputs, sub, idx_list):
if not isinstance(node.inputs[0].type, TensorType):
raise NotImplementedError()
def helper_c_code(node, name, inputs, outputs, sub, idx_list,
c_prefix="PyArray",
update_flags=("PyArray_UpdateFlags(xview,"
" NPY_ARRAY_C_CONTIGUOUS|"
"NPY_ARRAY_F_CONTIGUOUS);"),
set_data='PyArray_set_data',
set_dim='PyArray_set_dim',
set_stride='PyArray_set_stride',
strides_mul=1,
):
"""The parameters c_prefix, update_flags, set_data, set_dim,
set_stride and strides_mul are there to allow reusing this
function on PyArray and CudaNdarray object.
"""
#
# two arrays are created in C code:
# is_slice: len == ndim, 0 means int, 1 means slice
......@@ -4029,6 +4041,10 @@ class Subtensor(Op):
z, = outputs
rval = """
#define PyArray_set_dim(obj, idx, d) PyArray_DIMS(obj)[idx]=d
#define PyArray_set_stride(obj, idx, d) PyArray_STRIDES(obj)[idx]=d
#define PyArray_set_data(obj, ptr, base) PyArray_BYTES(obj)=ptr
// The subtensor is created by iterating over the dimensions
// and updating stride, shape, and data pointers
......@@ -4039,22 +4055,30 @@ class Subtensor(Op):
int inner_ii = 0; // the current dimension of zview
int outer_ii = 0; // current dimension of z
if ((PyArray_DIMS(xview) == PyArray_DIMS(%(x)s))
&& (PyArray_DIMS(%(x)s) != NULL))
char* ptr = (char*) %(c_prefix)s_BYTES(xview);
if ((%(c_prefix)s_DIMS(xview) == %(c_prefix)s_DIMS(%(x)s))
&& (%(c_prefix)s_DIMS(%(x)s) != NULL))
{
PyErr_Format(PyExc_ValueError, "x and xview"
"(with %%d dims) have the same dimensions"
" pointers: %%p and %%p",
PyArray_NDIM(%(x)s), PyArray_DIMS(xview), PyArray_DIMS(%(x)s));
%(c_prefix)s_NDIM(%(x)s),
%(c_prefix)s_DIMS(xview),
%(c_prefix)s_DIMS(%(x)s));
Py_XDECREF(xview);
%(fail)s;
}
if (PyArray_STRIDES(xview) == PyArray_STRIDES(%(x)s)
&& (PyArray_DIMS(%(x)s) != NULL))
if (%(c_prefix)s_STRIDES(xview) == %(c_prefix)s_STRIDES(%(x)s)
&& (%(c_prefix)s_DIMS(%(x)s) != NULL))
{
PyErr_Format(PyExc_ValueError, "x and xview"
"(with %%d dims) have the same strides"
" pointers: %%p and %%p",
PyArray_NDIM(%(x)s), PyArray_STRIDES(xview), PyArray_STRIDES(%(x)s));
%(c_prefix)s_NDIM(%(x)s),
%(c_prefix)s_STRIDES(xview),
%(c_prefix)s_STRIDES(%(x)s));
Py_XDECREF(xview);
%(fail)s;
}
......@@ -4062,7 +4086,7 @@ class Subtensor(Op):
{
if (is_slice[outer_ii])
{
npy_intp length = PyArray_DIMS(%(x)s)[outer_ii];
npy_intp length = %(c_prefix)s_DIMS(%(x)s)[outer_ii];
npy_intp slicelength;
npy_intp start = subtensor_spec[spec_pos+0];
npy_intp stop = subtensor_spec[spec_pos+1];
......@@ -4079,6 +4103,7 @@ class Subtensor(Op):
Py_DECREF(xview);
PyErr_Format(PyExc_ValueError,
"slice step cannot be zero");
Py_XDECREF(xview);
%(fail)s;
}
......@@ -4126,9 +4151,12 @@ class Subtensor(Op):
}
assert (slicelength <= length);
xview->data += PyArray_STRIDES(%(x)s)[outer_ii] * start;
PyArray_DIMS(xview)[inner_ii] = slicelength;
PyArray_STRIDES(xview)[inner_ii] = PyArray_STRIDES(%(x)s)[outer_ii] * step;
ptr += %(c_prefix)s_STRIDES(%(x)s)[outer_ii] * start *
%(strides_mul)s;
%(set_dim)s(xview, inner_ii, slicelength);
%(set_stride)s(xview, inner_ii,
%(c_prefix)s_STRIDES(%(x)s)[outer_ii] * step);
inner_ii += 1;
spec_pos += 3;
......@@ -4136,47 +4164,54 @@ class Subtensor(Op):
else // tuple coord `outer_ii` is an int
{
int idx = subtensor_spec[spec_pos];
if (idx < 0) idx += PyArray_DIMS(%(x)s)[outer_ii];
if (idx < 0) idx += %(c_prefix)s_DIMS(%(x)s)[outer_ii];
if (idx >= 0)
{
if (idx < PyArray_DIMS(%(x)s)[outer_ii])
if (idx < %(c_prefix)s_DIMS(%(x)s)[outer_ii])
{
xview->data += PyArray_STRIDES(%(x)s)[outer_ii] * idx;
ptr += %(c_prefix)s_STRIDES(%(x)s)[outer_ii] * idx *
%(strides_mul)s;
}
else
{
PyErr_Format(PyExc_IndexError,"index out of bounds");
Py_XDECREF(xview);
%(fail)s;
}
}
else
{
PyErr_Format(PyExc_IndexError,"index out of bounds");
Py_XDECREF(xview);
%(fail)s;
}
spec_pos += 1;
}
}
assert (inner_ii <= PyArray_NDIM(xview));
while (inner_ii < PyArray_NDIM(xview))
%(set_data)s(xview, ptr, (PyObject*)NULL);
assert (inner_ii <= %(c_prefix)s_NDIM(xview));
while (inner_ii < %(c_prefix)s_NDIM(xview))
{
assert (outer_ii < PyArray_NDIM(%(x)s));
PyArray_DIMS(xview)[inner_ii] = PyArray_DIMS(%(x)s)[outer_ii];
PyArray_STRIDES(xview)[inner_ii] = PyArray_STRIDES(%(x)s)[outer_ii];
assert (outer_ii < %(c_prefix)s_NDIM(%(x)s));
%(set_dim)s(xview, inner_ii, %(c_prefix)s_DIMS(%(x)s)[outer_ii]);
%(set_stride)s(xview, inner_ii, %(c_prefix)s_STRIDES(%(x)s)[outer_ii]);
inner_ii += 1;
outer_ii += 1;
}
PyArray_UpdateFlags(xview, NPY_ARRAY_C_CONTIGUOUS|NPY_F_CONTIGUOUS);
%(update_flags)s
""" % locals()
# print rval
return rval
@staticmethod
def helper_c_code_cache_version():
return (4,)
return (5,)
def c_code(self, node, name, inputs, outputs, sub): # DEBUG
if not isinstance(node.inputs[0].type, TensorType):
raise NotImplementedError()
x = inputs[0]
z, = outputs
view_ndim = node.outputs[0].ndim
......@@ -4219,7 +4254,7 @@ class Subtensor(Op):
# have a versioned version of this op's C code.
if len(hv) == 0:
return ()
return (1, hv)
return (2, hv)
def R_op(self, inputs, eval_points):
# Subtensor is not differentiable wrt to its indices, therefore we
......@@ -4479,6 +4514,8 @@ class IncSubtensor(Op):
out[0] = x
def c_code(self, node, name, inputs, outputs, sub): # DEBUG
if not isinstance(node.inputs[0].type, TensorType):
raise NotImplementedError()
if self.inplace: # convert bool to int
inplace = 1
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论