Design Article
Tell us What You Think
We want to know what you thought about this Design. Let us know by adding a comment.
Flexible and novel partitioning strategy for hierarchical design
Gurinder Singh Baghria, Kushagra Khorwal, Naveen Kumar - Freescale
12/27/2012 11:10 AM EST
Proposed approach
The solution to these limitations lies in adopting the flexible and novel partitioning strategy for hierarchically designs outlined in this paper. The attribute of the partition will be defined as COVER cell. Placement and routing of partitions is reclosed in ECO (Engineering Change Order) mode without any change in top-level placement. The proposed closure technique is subdivided into top-level (placement frozen) and block-level (ECO mode) implementation.
Reconsider Case 1 block-level implementation. Congestion is seen over the bottom side of partition A, when merged at the top. In order to resolve this congestion, the designer has to manipulate partition A to free some resources in terms of area for top. As discussed, shrinking the size of the partition will reset the entire block closure activities and need extra resources for re-closure. The proposed approach is based upon virtual shrinking of the block size. This means getting area from the blocks and allocating the same space for top in congested regions without changing actual block width/height and without major changes in block-level placement and routing. This can be done by generating empty area in the block, using placement and routing blockages where no standard cells are placed and no signals are routed.
Assume the floorplan for partition A is similar to that shown in figure 4.
Next: Case 2
The solution to these limitations lies in adopting the flexible and novel partitioning strategy for hierarchically designs outlined in this paper. The attribute of the partition will be defined as COVER cell. Placement and routing of partitions is reclosed in ECO (Engineering Change Order) mode without any change in top-level placement. The proposed closure technique is subdivided into top-level (placement frozen) and block-level (ECO mode) implementation.
Reconsider Case 1 block-level implementation. Congestion is seen over the bottom side of partition A, when merged at the top. In order to resolve this congestion, the designer has to manipulate partition A to free some resources in terms of area for top. As discussed, shrinking the size of the partition will reset the entire block closure activities and need extra resources for re-closure. The proposed approach is based upon virtual shrinking of the block size. This means getting area from the blocks and allocating the same space for top in congested regions without changing actual block width/height and without major changes in block-level placement and routing. This can be done by generating empty area in the block, using placement and routing blockages where no standard cells are placed and no signals are routed.
Assume the floorplan for partition A is similar to that shown in figure 4.
Figure 4: Case study floorplan for partition A (before and after implementation)
Due
to pin placement along the bottom side of partition A, there is little
to do there. In order to get some area from partition A, the only option
left is on the top side of the partition where memories and some logic
cell are placed. To get area from top, establish a placement and routing
blockage at the top side in partition A, and move the memories out from
these blockages (shifting down as shown in figure 4). Do the placement
and routing in ECO mode so that only the placement and routing of the
standard cells that are in the blockage region get re-placed and
re-routed rather than performing complete closure from scratch. This
will generate the white space at top side, which can be viewed as
virtually shrinking the block from Y to Y’= Y - delta Y from the top
side.Next: Case 2
|
|
|
|
|
Navigate to related information