dpctl.tensor¶
Data Parallel Tensor Collection
dpctl.tensor is an experimental collection of tensor implementations
that will implement future Python data API
(https://data-apis.github.io/array-api/latest/).
Available tensor implementations:
numpy_usm_shared:Provides a
numpy.ndarraysub-class whose underlying memory buffer is allocated with a USM shared memory allocator.
-
class
dpctl.tensor.usm_ndarray¶ - usm_ndarray(
shape, dtype=”|f8”, strides=None, buffer=’device’, offset=0, order=’C’, buffer_ctor_kwargs=dict()
)
See
dpctl.memory.MemoryUSMSharedfor allowed keyword arguments.buffer can be ‘shared’, ‘host’, ‘device’ to allocate new device memory by calling respective constructor with the specified buffer_ctor_kwrds; buffer can be an instance of
dpctl.memory.MemoryUSMShared,dpctl.memory.MemoryUSMDevice, ordpctl.memory.MemoryUSMHost; buffer can also be another usm_ndarray instance, in which case its underlying MemoryUSM* buffer is used for buffer.-
T¶
-
base_¶
-
device¶ Returns data-API object representing residence of the array data.
-
dtype¶ Returns NumPy’s dtype corresponding to the type of the array elements.
-
flags¶ Currently returns integer whose bits correspond to the flags.
-
flags_¶
-
imag¶
-
itemsize¶ Size of array element in bytes.
-
nbytes¶ Total bytes consumed by the elements of the array.
-
nd_¶
-
ndim¶ Gives the number of indices needed to address elements of this array.
-
real¶
-
shape¶ Elements of the shape tuple give the lengths of the respective array dimensions.
-
size¶ Number of elements in the array.
-
strides¶ Returns memory displacement in array elements, upon unit change of respective index.
E.g. for strides (s1, s2, s3) and multi-index (i1, i2, i3)
a[i1, i2, i3] == (&a[0,0,0])[ s1*s1 + s2*i2 + s3*i3]
-
sycl_context¶ Returns dpctl.SyclContext object to which USM data is bound.
-
sycl_device¶ Returns dpctl.SyclDevice object on which USM data was allocated.
-
sycl_queue¶ Returns dpctl.SyclQueue object associated with USM data.
-
to_device()¶ Transfer array to target device
-
typenum_¶
-
usm_data¶ Gives USM memory object underlying usm_array instance.
-
usm_type¶ USM type of underlying memory. Can be ‘device’, ‘shared’, or ‘host’.
See: https://docs.oneapi.com/versions/latest/dpcpp/iface/usm.html