Merge branch 'master' into cudnn-ctc

.flake8.cython

@@ -1,4 +1,4 @@
 [flake8]
 filename = *.pyx,*.px*
 exclude = .eggs,*.egg,build,docs,.git
-ignore = E225,E226,E227,E999
+ignore = W503,W504,E225,E226,E227,E999

.github/ISSUE_TEMPLATE.md

@@ -0,0 +1,14 @@
+Thank you for using and contributing to CuPy!
+
+If your issue is a **request for support in using CuPy**, please post it on [Stack Overflow](https://stackoverflow.com/questions/tagged/cupy) or Google Groups ([en](https://groups.google.com/forum/#!forum/cupy) / [ja](https://groups.google.com/forum/#!forum/cupy-ja)).
+Developer team members are monitoring questions on these channels.
+
+If it is a bug report, please include the following information:
+
+* Conditions (you can just paste the output of `python -c 'import cupy; cupy.show_config()'`)
+  - CuPy version
+  - OS/Platform
+  - CUDA version
+  - cuDNN/NCCL version (if applicable)
+* Code to reproduce
+* Error messages, stack traces, or logs

.travis.yml

@@ -24,7 +24,7 @@ script:
   - python -Werror::DeprecationWarning -Wignore::DeprecationWarning:site -m compileall -f -q cupy cupyx examples tests docs
   - pip install scipy
   - pushd docs
-  - make html
+  - SPHINXOPTS=-W make html
   - popd
 
 sudo: false

CODE_OF_CONDUCT.md

@@ -0,0 +1,7 @@
+# Chainer Code of Conduct
+
+CuPy follows the [NumFOCUS Code of Conduct][homepage] available at https://numfocus.org/code-of-conduct.
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be reported by contacting the project team at chainer@preferred.jp. 
+
+[homepage]: https://numfocus.org/

MANIFEST.in

@@ -1,4 +1,4 @@
-recursive-include cupy *.h
+recursive-include cupy *.h *.hpp
 recursive-exclude cupy *.pyx *.pxd *.pxi
 include cupy_setup_build.py
 recursive-include install *.py

README.md

@@ -1,4 +1,4 @@
-<div align="center"><img src="docs/image/cupy_logo_1000px.png" width="400"/></div>
+<div align="center"><img src="https://raw.githubusercontent.com/cupy/cupy/master/docs/image/cupy_logo_1000px.png" width="400"/></div>
 
 # CuPy : NumPy-like API accelerated with CUDA
 
@@ -38,6 +38,9 @@ $ pip install cupy-cuda91
 (Binary Package for CUDA 9.2)
 $ pip install cupy-cuda92
 
+(Binary Package for CUDA 10.0)
+$ pip install cupy-cuda100
+
 (Source Package)
 $ pip install cupy
 ```

cupy/__init__.py

@@ -58,6 +58,7 @@ def is_available():
 from cupy import statistics  # NOQA
 from cupy import testing  # NOQA  # NOQA
 from cupy import util  # NOQA
+from cupy import lib  # NOQA
 
 
 # import class and function
@@ -387,6 +388,8 @@ def binary_repr(num, width=None):
 from cupy.indexing.indexing import diagonal  # NOQA
 from cupy.indexing.indexing import take  # NOQA
 
+from cupy.indexing.insert import place  # NOQA
+from cupy.indexing.insert import put  # NOQA
 from cupy.indexing.insert import fill_diagonal  # NOQA
 # -----------------------------------------------------------------------------
 # Input and output
@@ -478,6 +481,7 @@ def isscalar(num):
 from cupy.math.trigonometric import radians  # NOQA
 from cupy.math.trigonometric import sin  # NOQA
 from cupy.math.trigonometric import tan  # NOQA
+from cupy.math.trigonometric import unwrap  # NOQA
 
 from cupy.math.hyperbolic import arccosh  # NOQA
 from cupy.math.hyperbolic import arcsinh  # NOQA
@@ -498,6 +502,7 @@ def isscalar(num):
 from cupy.math.sumprod import sum  # NOQA
 from cupy.math.sumprod import cumprod  # NOQA
 from cupy.math.sumprod import cumsum  # NOQA
+from cupy.math.sumprod import diff  # NOQA
 from cupy.math.window import blackman  # NOQA
 from cupy.math.window import hamming  # NOQA
 from cupy.math.window import hanning  # NOQA
@@ -627,7 +632,7 @@ def isscalar(num):
 from cupy.ext.scatter import scatter_add  # NOQA
 
 
-def asnumpy(a, stream=None):
+def asnumpy(a, stream=None, order='C'):
     """Returns an array on the host memory from an arbitrary source array.
 
     Args:
@@ -636,15 +641,17 @@ def asnumpy(a, stream=None):
             the device-to-host copy runs asynchronously. Otherwise, the copy is
             synchronous. Note that if ``a`` is not a :class:`cupy.ndarray`
             object, then this argument has no effect.
-
+        order ({'C', 'F', 'A'}): The desired memory layout of the host
+            array. When ``order`` is 'A', it uses 'F' if ``a`` is
+            fortran-contiguous and 'C' otherwise.
     Returns:
         numpy.ndarray: Converted array on the host memory.
 
     """
     if isinstance(a, ndarray):
-        return a.get(stream=stream)
+        return a.get(stream=stream, order=order)
     else:
-        return numpy.asarray(a)
+        return numpy.asarray(a, order=order)
 
 
 _cupy = sys.modules[__name__]
@@ -675,7 +682,8 @@ def get_array_module(*args):
     """
     for arg in args:
         if isinstance(arg, (ndarray, sparse.spmatrix,
-                            cupy.core.fusion.FusionVarPython)):
+                            cupy.core.fusion._FusionVarScalar,
+                            cupy.core.fusion._FusionVarArray)):
             return _cupy
     return numpy
 

cupy/_version.py

@@ -1 +1 @@
-__version__ = '6.0.0a1'
+__version__ = '6.0.0rc1'

cupy/core/__init__.py

@@ -3,59 +3,59 @@
 
 
 # import class and function
+from cupy.core._errors import _AxisError  # NOQA
 from cupy.core._kernel import create_reduction_func  # NOQA
 from cupy.core._kernel import create_ufunc  # NOQA
 from cupy.core._kernel import ElementwiseKernel  # NOQA
 from cupy.core._kernel import ReductionKernel  # NOQA
 from cupy.core._kernel import ufunc  # NOQA
-from cupy.core.core import absolute  # NOQA
-from cupy.core.core import add  # NOQA
-from cupy.core.core import angle  # NOQA
+from cupy.core._routines_manipulation import array_split  # NOQA
+from cupy.core._routines_manipulation import broadcast  # NOQA
+from cupy.core._routines_manipulation import broadcast_to  # NOQA
+from cupy.core._routines_manipulation import concatenate_method  # NOQA
+from cupy.core._routines_manipulation import moveaxis  # NOQA
+from cupy.core._routines_manipulation import rollaxis  # NOQA
+from cupy.core._routines_manipulation import size  # NOQA'
+from cupy.core._routines_math import absolute  # NOQA
+from cupy.core._routines_math import add  # NOQA
+from cupy.core._routines_math import angle  # NOQA
+from cupy.core._routines_math import conj  # NOQA
+from cupy.core._routines_math import divide  # NOQA
+from cupy.core._routines_math import floor_divide  # NOQA
+from cupy.core._routines_math import imag  # NOQA
+from cupy.core._routines_math import multiply  # NOQA
+from cupy.core._routines_math import negative  # NOQA
+from cupy.core._routines_math import power  # NOQA
+from cupy.core._routines_math import real  # NOQA
+from cupy.core._routines_math import remainder  # NOQA
+from cupy.core._routines_math import sqrt  # NOQA
+from cupy.core._routines_math import subtract  # NOQA
+from cupy.core._routines_math import true_divide  # NOQA
+from cupy.core._routines_statistics import nanmax  # NOQA
+from cupy.core._routines_statistics import nanmin  # NOQA
 from cupy.core.core import array  # NOQA
-from cupy.core.core import array_split  # NOQA
 from cupy.core.core import ascontiguousarray  # NOQA
 from cupy.core.core import asfortranarray  # NOQA
 from cupy.core.core import bitwise_and  # NOQA
 from cupy.core.core import bitwise_or  # NOQA
 from cupy.core.core import bitwise_xor  # NOQA
-from cupy.core.core import broadcast  # NOQA
-from cupy.core.core import broadcast_to  # NOQA
-from cupy.core.core import concatenate_method  # NOQA
-from cupy.core.core import conj  # NOQA
 from cupy.core.core import create_comparison  # NOQA
-from cupy.core.core import divide  # NOQA
 from cupy.core.core import divmod  # NOQA
 from cupy.core.core import dot  # NOQA
 from cupy.core.core import elementwise_copy  # NOQA
 from cupy.core.core import elementwise_copy_where  # NOQA
 from cupy.core.core import equal  # NOQA
-from cupy.core.core import floor_divide  # NOQA
 from cupy.core.core import greater  # NOQA
 from cupy.core.core import greater_equal  # NOQA
-from cupy.core.core import imag  # NOQA
 from cupy.core.core import invert  # NOQA
 from cupy.core.core import left_shift  # NOQA
 from cupy.core.core import less  # NOQA
 from cupy.core.core import less_equal  # NOQA
 from cupy.core.core import matmul  # NOQA
-from cupy.core.core import moveaxis  # NOQA
-from cupy.core.core import multiply  # NOQA
-from cupy.core.core import nanmax  # NOQA
-from cupy.core.core import nanmin  # NOQA
 from cupy.core.core import ndarray  # NOQA
-from cupy.core.core import negative  # NOQA
-from cupy.core.core import normalize_axis_tuple  # NOQA
 from cupy.core.core import not_equal  # NOQA
-from cupy.core.core import power  # NOQA
-from cupy.core.core import real  # NOQA
-from cupy.core.core import remainder  # NOQA
 from cupy.core.core import right_shift  # NOQA
-from cupy.core.core import rollaxis  # NOQA
-from cupy.core.core import size  # NOQA'
-from cupy.core.core import sqrt  # NOQA
-from cupy.core.core import subtract  # NOQA
 from cupy.core.core import tensordot_core  # NOQA
-from cupy.core.core import true_divide  # NOQA
 from cupy.core.dlpack import fromDlpack  # NOQA
 from cupy.core.internal import complete_slice  # NOQA
 from cupy.core.internal import get_size  # NOQA

cupy/core/_dtype.pyx

@@ -45,7 +45,7 @@ _init_dtype_dict()
 
 @cython.profile(False)
 cpdef get_dtype(t):
-    if isinstance(t, numpy.dtype):
+    if type(t) is numpy.dtype:  # Exact type check
         return t
     ret = _dtype_dict.get(t, None)
     if ret is None:
@@ -55,7 +55,7 @@ cpdef get_dtype(t):
 
 @cython.profile(False)
 cpdef tuple get_dtype_with_itemsize(t):
-    if isinstance(t, numpy.dtype):
+    if type(t) is numpy.dtype:  # Exact type check
         return t, t.itemsize
     ret = _dtype_dict.get(t, None)
     if ret is None:

cupy/core/_errors.py

@@ -0,0 +1,12 @@
+import numpy
+
+
+try:
+    _AxisError = numpy.AxisError
+except AttributeError:
+    class IndexOrValueError(IndexError, ValueError):
+
+        def __init__(self, *args, **kwargs):
+            super(IndexOrValueError, self).__init__(*args, **kwargs)
+
+    _AxisError = IndexOrValueError