You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Got fixed when the key for the USMNdArray type was updated to include the queue. The reproducer now raises an ExecutionQueueInferenceError with following traceback:
Traceback (most recent call last):
File "/localdisk/work/diptorup/devel/numba-dpex/driver.py", line 23, in<module>
sum_kernel[dpex.Range(1)](a, b, c)
File "/localdisk/work/diptorup/devel/numba-dpex/numba_dpex/core/kernel_interface/dispatcher.py", line 412, in __call__
ty_queue = determine_kernel_launch_queue(
File "/localdisk/work/diptorup/devel/numba-dpex/numba_dpex/core/kernel_interface/utils.py", line 72, in determine_kernel_launch_queue
raise ExecutionQueueInferenceError(
numba_dpex.core.exceptions.ExecutionQueueInferenceError: Execution queue for kernel "sum_kernel" could be deduced using compute follows data programming model. The usm_ndarray arguments 0,1,2 were not allocated on the same queue.
The following code should raise a
ComputeFollowsDataInferenceError
, however such an exception is not raised:Inside dispatcher when the type of the input arguments is inferred I see that all the args are inferred as
usm_ndarray
type with the same queue.The text was updated successfully, but these errors were encountered: