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
Currently, although multiple bel pins can map to a single cell pin, there is no way to specify that the site router is free to choose any of them (except for LUTs which are implicitly assumed to be permutable based on the LUT bel info). If one cell pin maps to multiple bel pins at the moment, all of them will be routed (as is needed for BRAM, for example).
I see three ways of solving this:
not supporting it, and adding virtual site pips to implement the choice of routing, with conversion necessary when generating a DCP or otherwise interfacing with the Xilinx ecosystem
adding it as an explicit special case in its own right
treating it as a special case of the parameter bel pin mapping rules (basically saying that the site router is free to change this parameter and hence the mapping as it likes).
The text was updated successfully, but these errors were encountered:
A similar situation exists with the CASC bel pin of Versal LUTs that can be used as a LUT input in some cases, which might create some interesting interactions with the current implicitly permutable LUTs...
Consider a Xilinx CARRY8. The
DI[i]
cell pins (i≠4) can map either to an 'aX' bel pin or a DIi bel pin:(
set_property LOCK_PINS {DI[1]:DI1} [get_cells carry_i]
)(
set_property LOCK_PINS {DI[1]:BX} [get_cells carry_i]
)Currently, although multiple bel pins can map to a single cell pin, there is no way to specify that the site router is free to choose any of them (except for LUTs which are implicitly assumed to be permutable based on the LUT bel info). If one cell pin maps to multiple bel pins at the moment, all of them will be routed (as is needed for BRAM, for example).
I see three ways of solving this:
The text was updated successfully, but these errors were encountered: