Release 0.11.2

CHANGELOG.md

@@ -4,6 +4,16 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+## [0.11.2] - 11/29/2021
+
+### Added
+- Extending `dpctl.device_context` with nested contexts (#678)
+
+## Fixed
+- Fixed issue #649 about incorrect behavior of `.T` method on sliced arrays (#653)
+
 ## [0.11.1] - 11/10/2021
 
 ### Changed

dpctl/__init__.py

@@ -61,6 +61,7 @@
     get_current_queue,
     get_num_activated_queues,
     is_in_device_context,
+    nested_context_factories,
     set_global_queue,
 )
 
@@ -111,6 +112,7 @@
     "get_current_queue",
     "get_num_activated_queues",
     "is_in_device_context",
+    "nested_context_factories",
     "set_global_queue",
 ]
 __all__ += [

dpctl/_sycl_queue_manager.pyx

@@ -19,7 +19,7 @@
 # cython: linetrace=True
 
 import logging
-from contextlib import contextmanager
+from contextlib import ExitStack, contextmanager
 
 from .enum_types import backend_type, device_type
 
@@ -210,6 +210,22 @@ cpdef get_current_backend():
     return _mgr.get_current_backend()
 
 
+nested_context_factories = []
+
+
+def _get_nested_contexts(ctxt):
+    _help_numba_dppy()
+    return (factory(ctxt) for factory in nested_context_factories)
+
+
+def _help_numba_dppy():
+    """Import numba-dppy for registering nested contexts"""
+    try:
+        import numba_dppy
+    except Exception:
+        pass
+
+
 @contextmanager
 def device_context(arg):
     """
@@ -222,6 +238,9 @@ def device_context(arg):
     the context manager's scope. The yielded queue is removed as the currently
     usable queue on exiting the context manager.
 
+    You can register context factory in the list of factories.
+    This context manager uses context factories to create and activate nested contexts.
+
     Args:
 
         queue_str (str) : A string corresponding to the DPC++ filter selector.
@@ -243,11 +262,26 @@ def device_context(arg):
             with dpctl.device_context("level0:gpu:0"):
                 pass
 
+        The following example registers nested context factory:
+
+        .. code-block:: python
+
+            import dctl
+
+            def factory(sycl_queue):
+                ...
+                return context
+
+            dpctl.nested_context_factories.append(factory)
+
     """
     ctxt = None
     try:
         ctxt = _mgr._set_as_current_queue(arg)
-        yield ctxt
+        with ExitStack() as stack:
+            for nested_context in _get_nested_contexts(ctxt):
+                stack.enter_context(nested_context)
+            yield ctxt
     finally:
         # Code to release resource
         if ctxt:

dpctl/tensor/_usmarray.pyx

@@ -1014,31 +1014,62 @@ cdef usm_ndarray _real_view(usm_ndarray ary):
     """
     View into real parts of a complex type array
     """
-    cdef usm_ndarray r = ary._clone()
+    cdef int r_typenum_ = -1
+    cdef usm_ndarray r = None
+    cdef Py_ssize_t offset_elems = 0
+
     if (ary.typenum_ == UAR_CFLOAT):
-        r.typenum_ = UAR_FLOAT
+        r_typenum_ = UAR_FLOAT
     elif (ary.typenum_ == UAR_CDOUBLE):
-        r.typenum_ = UAR_DOUBLE
+        r_typenum_ = UAR_DOUBLE
     else:
         raise InternalUSMArrayError(
             "_real_view call on array of non-complex type.")
+
+    offset_elems = ary.get_offset() * 2
+    r = usm_ndarray.__new__(
+        usm_ndarray,
+        _make_int_tuple(ary.nd_, ary.shape_),
+        dtype=_make_typestr(r_typenum_),
+        strides=tuple(2 * si for si in ary.strides),
+        buffer=ary.base_,
+        offset=offset_elems,
+        order=('C' if (ary.flags_ & USM_ARRAY_C_CONTIGUOUS) else 'F')
+    )
+    r.flags_ = ary.flags_
+    r.array_namespace_ = ary.array_namespace_
     return r
 
 
 cdef usm_ndarray _imag_view(usm_ndarray ary):
     """
     View into imaginary parts of a complex type array
     """
-    cdef usm_ndarray r = ary._clone()
+    cdef int r_typenum_ = -1
+    cdef usm_ndarray r = None
+    cdef Py_ssize_t offset_elems = 0
+
     if (ary.typenum_ == UAR_CFLOAT):
-        r.typenum_ = UAR_FLOAT
+        r_typenum_ = UAR_FLOAT
     elif (ary.typenum_ == UAR_CDOUBLE):
-        r.typenum_ = UAR_DOUBLE
+        r_typenum_ = UAR_DOUBLE
     else:
         raise InternalUSMArrayError(
-            "_real_view call on array of non-complex type.")
+            "_imag_view call on array of non-complex type.")
+
     # displace pointer to imaginary part
-    r.data_ = r.data_ + type_bytesize(r.typenum_)
+    offset_elems = 2 * ary.get_offset() + 1
+    r = usm_ndarray.__new__(
+        usm_ndarray,
+        _make_int_tuple(ary.nd_, ary.shape_),
+        dtype=_make_typestr(r_typenum_),
+        strides=tuple(2 * si for si in ary.strides),
+        buffer=ary.base_,
+        offset=offset_elems,
+        order=('C' if (ary.flags_ & USM_ARRAY_C_CONTIGUOUS) else 'F')
+    )
+    r.flags_ = ary.flags_
+    r.array_namespace_ = ary.array_namespace_
     return r
 
 
@@ -1054,7 +1085,8 @@ cdef usm_ndarray _transpose(usm_ndarray ary):
             _make_reversed_int_tuple(ary.nd_, ary.strides_)
             if (ary.strides_) else None),
         buffer=ary.base_,
-        order=('F' if (ary.flags_ & USM_ARRAY_C_CONTIGUOUS) else 'C')
+        order=('F' if (ary.flags_ & USM_ARRAY_C_CONTIGUOUS) else 'C'),
+        offset=ary.get_offset()
     )
     r.flags_ |= (ary.flags_ & USM_ARRAY_WRITEABLE)
     return r

dpctl/tests/test_sycl_queue_manager.py

@@ -17,6 +17,8 @@
 """Defines unit test cases for the SyclQueueManager class.
 """
 
+import contextlib
+
 import pytest
 
 import dpctl
@@ -156,3 +158,73 @@ def test_get_current_backend():
         dpctl.set_global_queue("gpu")
     elif has_cpu():
         dpctl.set_global_queue("cpu")
+
+
+def test_nested_context_factory_is_empty_list():
+    assert isinstance(dpctl.nested_context_factories, list)
+    assert not dpctl.nested_context_factories
+
+
+@contextlib.contextmanager
+def _register_nested_context_factory(factory):
+    dpctl.nested_context_factories.append(factory)
+    try:
+        yield
+    finally:
+        dpctl.nested_context_factories.remove(factory)
+
+
+def test_register_nested_context_factory_context():
+    def factory():
+        pass
+
+    with _register_nested_context_factory(factory):
+        assert factory in dpctl.nested_context_factories
+
+    assert isinstance(dpctl.nested_context_factories, list)
+    assert not dpctl.nested_context_factories
+
+
+@pytest.mark.skipif(not has_cpu(), reason="No OpenCL CPU queues available")
+def test_device_context_activates_nested_context():
+    in_context = False
+    factory_called = False
+
+    @contextlib.contextmanager
+    def context():
+        nonlocal in_context
+        old, in_context = in_context, True
+        yield
+        in_context = old
+
+    def factory(_):
+        nonlocal factory_called
+        factory_called = True
+        return context()
+
+    with _register_nested_context_factory(factory):
+        assert not factory_called
+        assert not in_context
+
+        with dpctl.device_context("opencl:cpu:0"):
+            assert factory_called
+            assert in_context
+
+        assert not in_context
+
+
+@pytest.mark.skipif(not has_cpu(), reason="No OpenCL CPU queues available")
+@pytest.mark.parametrize(
+    "factory, exception, match",
+    [
+        (True, TypeError, "object is not callable"),
+        (lambda x: None, AttributeError, "no attribute '__exit__'"),
+    ],
+)
+def test_nested_context_factory_exception_if_wrong_factory(
+    factory, exception, match
+):
+    with pytest.raises(exception, match=match):
+        with _register_nested_context_factory(factory):
+            with dpctl.device_context("opencl:cpu:0"):
+                pass

dpctl/tests/test_usm_ndarray_ctor.py

@@ -841,3 +841,25 @@ def test_reshape():
         dpt.reshape(Z, Z.shape, order="invalid")
     W = dpt.reshape(Z, (-1,), order="C")
     assert W.shape == (Z.size,)
+
+
+def test_transpose():
+    n, m = 2, 3
+    X = dpt.usm_ndarray((n, m), "f4")
+    Xnp = np.arange(n * m, dtype="f4").reshape((n, m))
+    X[:] = Xnp
+    assert np.array_equal(dpt.to_numpy(X.T), Xnp.T)
+    assert np.array_equal(dpt.to_numpy(X[1:].T), Xnp[1:].T)
+
+
+def test_real_imag_views():
+    n, m = 2, 3
+    X = dpt.usm_ndarray((n, m), "c8")
+    Xnp_r = np.arange(n * m, dtype="f4").reshape((n, m))
+    Xnp_i = np.arange(n * m, 2 * n * m, dtype="f4").reshape((n, m))
+    Xnp = Xnp_r + 1j * Xnp_i
+    X[:] = Xnp
+    assert np.array_equal(dpt.to_numpy(X.real), Xnp.real)
+    assert np.array_equal(dpt.to_numpy(X.imag), Xnp.imag)
+    assert np.array_equal(dpt.to_numpy(X[1:].real), Xnp[1:].real)
+    assert np.array_equal(dpt.to_numpy(X[1:].imag), Xnp[1:].imag)