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
To define a shared memory link between two cores, the top level memory should be exactly the same. This means that also the served dimensions should be the same, and in case the top level memory serves both dimensions of the 2D array, the top level memory of the 1D array must also have 2 served dimensions.
This can easily be mitigated by defining the 1D array as a 2D array [D1, D2] with sizes [128, 1], and unrolling an arbitrary dimension at D2 (e.g. D2: (B, 1)) to prevent multiple runs of the spatial mapping generator.
This approach is cumbersome and confusing for first time users.
The text was updated successfully, but these errors were encountered:
To define a shared memory link between two cores, the top level memory should be exactly the same. This means that also the served dimensions should be the same, and in case the top level memory serves both dimensions of the 2D array, the top level memory of the 1D array must also have 2 served dimensions.
This can easily be mitigated by defining the 1D array as a 2D array
[D1, D2]
with sizes[128, 1]
, and unrolling an arbitrary dimension atD2
(e.g.D2: (B, 1)
) to prevent multiple runs of the spatial mapping generator.This approach is cumbersome and confusing for first time users.
The text was updated successfully, but these errors were encountered: