__array_namespace_info__ returns complex128 dtype when no fp64 #1977
Description
It seems dpctl only checks if float64 is supported based on device capabilities when building a dictionary of all Array API data types:
import dpctl, dpctl.tensor as dpt
dpctl.__version__
# Out: '0.19.0dev0+480.g360ce45dc6'
info = dpnp.__array_namespace_info__()
info.dtypes()
# Out:
# {'bool': dtype('bool'),
# 'float32': dtype('float32'),
# 'complex64': dtype('complex64'),
# 'complex128': dtype('complex128'),
# 'int8': dtype('int8'),
# 'int16': dtype('int16'),
# 'int32': dtype('int32'),
# 'int64': dtype('int64'),
# 'uint8': dtype('uint8'),
# 'uint16': dtype('uint16'),
# 'uint32': dtype('uint32'),
# 'uint64': dtype('uint64')}
# no fp64 support by default device:
dpctl.select_default_device().has_aspect_fp64
# Out: FalseAs a result float64 is filtered out but complex128 is present in the output dictionary, which looks faulty.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
Something isn't workingSomething isn't working