Merge branch 'master' into add-dtypes-and-infos

Author: oleksandr-pavlyk

.flake8

@@ -25,6 +25,7 @@ per-file-ignores =
     dpctl/program/_program.pyx: E999, E225, E226, E227
     dpctl/tensor/_usmarray.pyx: E999, E225, E226, E227
     dpctl/tensor/_dlpack.pyx: E999, E225, E226, E227
+    dpctl/tensor/_flags.pyx: E999, E225, E226, E227
     dpctl/tensor/numpy_usm_shared.py: F821
     dpctl/tests/_cython_api.pyx: E999, E225, E227, E402
     dpctl/utils/_compute_follows_data.pyx: E999, E225, E227

dpctl/memory/_memory.pxd

@@ -75,6 +75,6 @@ cdef public api class MemoryUSMHost(_Memory) [object PyMemoryUSMHostObject,
     pass
 
 
-cdef public class MemoryUSMDevice(_Memory) [object PyMemoryUSMDeviceObject,
+cdef public api class MemoryUSMDevice(_Memory) [object PyMemoryUSMDeviceObject,
                                             type PyMemoryUSMDeviceType]:
     pass

dpctl/tensor/__init__.py

@@ -33,6 +33,7 @@
     full,
     full_like,
     linspace,
+    meshgrid,
     ones,
     ones_like,
     tril,
@@ -127,4 +128,5 @@
     "finfo",
     "can_cast",
     "result_type",
+    "meshgrid",
 ]

dpctl/tensor/_copy_utils.py

@@ -261,18 +261,18 @@ def copy(usm_ary, order="K"):
     elif order == "F":
         copy_order = order
     elif order == "A":
-        if usm_ary.flags & 2:
+        if usm_ary.flags.f_contiguous:
             copy_order = "F"
     elif order == "K":
-        if usm_ary.flags & 2:
+        if usm_ary.flags.f_contiguous:
             copy_order = "F"
     else:
         raise ValueError(
             "Unrecognized value of the order keyword. "
             "Recognized values are 'A', 'C', 'F', or 'K'"
         )
-    c_contig = usm_ary.flags & 1
-    f_contig = usm_ary.flags & 2
+    c_contig = usm_ary.flags.c_contiguous
+    f_contig = usm_ary.flags.f_contiguous
     R = dpt.usm_ndarray(
         usm_ary.shape,
         dtype=usm_ary.dtype,
@@ -325,8 +325,8 @@ def astype(usm_ary, newdtype, order="K", casting="unsafe", copy=True):
                 ary_dtype, newdtype, casting
             )
         )
-    c_contig = usm_ary.flags & 1
-    f_contig = usm_ary.flags & 2
+    c_contig = usm_ary.flags.c_contiguous
+    f_contig = usm_ary.flags.f_contiguous
     needs_copy = copy or not (ary_dtype == target_dtype)
     if not needs_copy and (order != "K"):
         needs_copy = (c_contig and order not in ["A", "C"]) or (
@@ -339,10 +339,10 @@ def astype(usm_ary, newdtype, order="K", casting="unsafe", copy=True):
         elif order == "F":
             copy_order = order
         elif order == "A":
-            if usm_ary.flags & 2:
+            if usm_ary.flags.f_contiguous:
                 copy_order = "F"
         elif order == "K":
-            if usm_ary.flags & 2:
+            if usm_ary.flags.f_contiguous:
                 copy_order = "F"
         else:
             raise ValueError(

dpctl/tensor/_ctors.py

@@ -133,9 +133,9 @@ def _asarray_from_usm_ndarray(
     #    sycl_queue is unchanged
     can_zero_copy = can_zero_copy and copy_q is usm_ndary.sycl_queue
     #    order is unchanged
-    c_contig = usm_ndary.flags & 1
-    f_contig = usm_ndary.flags & 2
-    fc_contig = usm_ndary.flags & 3
+    c_contig = usm_ndary.flags.c_contiguous
+    f_contig = usm_ndary.flags.f_contiguous
+    fc_contig = usm_ndary.flags.forc
     if can_zero_copy:
         if order == "C" and c_contig:
             pass
@@ -1130,7 +1130,7 @@ def tril(X, k=0):
     k = operator.index(k)
 
     # F_CONTIGUOUS = 2
-    order = "F" if (X.flags & 2) else "C"
+    order = "F" if (X.flags.f_contiguous) else "C"
 
     shape = X.shape
     nd = X.ndim
@@ -1171,7 +1171,7 @@ def triu(X, k=0):
     k = operator.index(k)
 
     # F_CONTIGUOUS = 2
-    order = "F" if (X.flags & 2) else "C"
+    order = "F" if (X.flags.f_contiguous) else "C"
 
     shape = X.shape
     nd = X.ndim
@@ -1198,3 +1198,61 @@ def triu(X, k=0):
         hev.wait()
 
     return res
+
+
+def meshgrid(*arrays, indexing="xy"):
+
+    """
+    meshgrid(*arrays, indexing="xy") -> list[usm_ndarray]
+
+    Creates list of `usm_ndarray` coordinate matrices from vectors.
+
+    Args:
+        arrays: arbitrary number of one-dimensional `USM_ndarray` objects.
+            If vectors are not of the same data type,
+            or are not one-dimensional, raises `ValueError.`
+        indexing: Cartesian (`xy`) or matrix (`ij`) indexing of output.
+            For a set of `n` vectors with lengths N0, N1, N2, ...
+            Cartesian indexing results in arrays of shape
+            (N1, N0, N2, ...)
+            matrix indexing results in arrays of shape
+            (n0, N1, N2, ...)
+            Default: `xy`.
+    """
+    ref_dt = None
+    ref_unset = True
+    for array in arrays:
+        if not isinstance(array, dpt.usm_ndarray):
+            raise TypeError(
+                f"Expected instance of dpt.usm_ndarray, got {type(array)}."
+            )
+        if array.ndim != 1:
+            raise ValueError("All arrays must be one-dimensional.")
+        if ref_unset:
+            ref_unset = False
+            ref_dt = array.dtype
+        else:
+            if not ref_dt == array.dtype:
+                raise ValueError(
+                    "All arrays must be of the same numeric data type."
+                )
+    if indexing not in ["xy", "ij"]:
+        raise ValueError(
+            "Unrecognized indexing keyword value, expecting 'xy' or 'ij.'"
+        )
+    n = len(arrays)
+    sh = (-1,) + (1,) * (n - 1)
+
+    res = []
+    if n > 1 and indexing == "xy":
+        res.append(dpt.reshape(arrays[0], (1, -1) + sh[2:], copy=True))
+        res.append(dpt.reshape(arrays[1], sh, copy=True))
+        arrays, sh = arrays[2:], sh[-2:] + sh[:-2]
+
+    for array in arrays:
+        res.append(dpt.reshape(array, sh, copy=True))
+        sh = sh[-1:] + sh[:-1]
+
+    output = dpt.broadcast_arrays(*res)
+
+    return output

dpctl/tensor/_flags.pyx

@@ -0,0 +1,111 @@
+#                       Data Parallel Control (dpctl)
+#
+#  Copyright 2020-2022 Intel Corporation
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+# distutils: language = c++
+# cython: language_level=3
+# cython: linetrace=True
+
+from libcpp cimport bool as cpp_bool
+
+from dpctl.tensor._usmarray cimport (
+    USM_ARRAY_C_CONTIGUOUS,
+    USM_ARRAY_F_CONTIGUOUS,
+    USM_ARRAY_WRITEABLE,
+    usm_ndarray,
+)
+
+
+cdef cpp_bool _check_bit(int flag, int mask):
+    return (flag & mask) == mask
+
+
+cdef class Flags:
+    """Helper class to represent flags of :class:`dpctl.tensor.usm_ndarray`."""
+    cdef int flags_
+    cdef usm_ndarray arr_
+
+    def __cinit__(self, usm_ndarray arr, int flags):
+        self.arr_ = arr
+        self.flags_ = flags
+
+    @property
+    def flags(self):
+        return self.flags_
+
+    @property
+    def c_contiguous(self):
+        return _check_bit(self.flags_, USM_ARRAY_C_CONTIGUOUS)
+
+    @property
+    def f_contiguous(self):
+        return _check_bit(self.flags_, USM_ARRAY_F_CONTIGUOUS)
+
+    @property
+    def writable(self):
+        return _check_bit(self.flags_, USM_ARRAY_WRITEABLE)
+
+    @property
+    def fc(self):
+        return (
+           _check_bit(self.flags_, USM_ARRAY_C_CONTIGUOUS)
+           and _check_bit(self.flags_, USM_ARRAY_F_CONTIGUOUS)
+        )
+
+    @property
+    def forc(self):
+        return (
+           _check_bit(self.flags_, USM_ARRAY_C_CONTIGUOUS)
+           or _check_bit(self.flags_, USM_ARRAY_F_CONTIGUOUS)
+        )
+
+    @property
+    def fnc(self):
+        return (
+           _check_bit(self.flags_, USM_ARRAY_C_CONTIGUOUS)
+           and not _check_bit(self.flags_, USM_ARRAY_F_CONTIGUOUS)
+        )
+
+    @property
+    def contiguous(self):
+        return self.forc
+
+    def __getitem__(self, name):
+        if name in ["C_CONTIGUOUS", "C"]:
+            return self.c_contiguous
+        elif name in ["F_CONTIGUOUS", "F"]:
+            return self.f_contiguous
+        elif name == "WRITABLE":
+            return self.writable
+        elif name == "FC":
+            return self.fc
+        elif name == "CONTIGUOUS":
+            return self.forc
+
+    def __repr__(self):
+        out = []
+        for name in "C_CONTIGUOUS", "F_CONTIGUOUS", "WRITABLE":
+            out.append("  {} : {}".format(name, self[name]))
+        return '\n'.join(out)
+
+    def __eq__(self, other):
+        cdef Flags other_
+        if isinstance(other, self.__class__):
+           other_ = <Flags>other
+           return self.flags_ == other_.flags_
+        elif isinstance(other, int):
+           return self.flags_ == <int>other
+        else:
+           return False

dpctl/tensor/_reshape.py

@@ -86,7 +86,9 @@ def reshape(X, newshape, order="C", copy=None):
         raise TypeError
     if not isinstance(newshape, (list, tuple)):
         newshape = (newshape,)
-    if order not in ["C", "F"]:
+    if order in "cfCF":
+        order = order.upper()
+    else:
         raise ValueError(
             f"Keyword 'order' not recognized. Expecting 'C' or 'F', got {order}"
         )

dpctl/tensor/_usmarray.pyx

@@ -33,6 +33,7 @@ from cpython.tuple cimport PyTuple_New, PyTuple_SetItem
 cimport dpctl as c_dpctl
 cimport dpctl.memory as c_dpmem
 cimport dpctl.tensor._dlpack as c_dlpack
+import dpctl.tensor._flags as _flags
 
 include "_stride_utils.pxi"
 include "_types.pxi"
@@ -503,9 +504,9 @@ cdef class usm_ndarray:
     @property
     def flags(self):
         """
-        Currently returns integer whose bits correspond to the flags.
+        Returns dpctl.tensor._flags object.
         """
-        return self.flags_
+        return _flags.Flags(self, self.flags_)
 
     @property
     def usm_type(self):
@@ -663,7 +664,7 @@ cdef class usm_ndarray:
                 strides=self.strides,
                 offset=self.get_offset()
             )
-            res.flags_ = self.flags
+            res.flags_ = self.flags.flags
             return res
         else:
             nbytes = self.usm_data.nbytes
@@ -678,7 +679,7 @@ cdef class usm_ndarray:
                 strides=self.strides,
                 offset=self.get_offset()
             )
-            res.flags_ = self.flags
+            res.flags_ = self.flags.flags
             return res
 
     def _set_namespace(self, mod):