Physical design automation system and method using hierarchical clusterization and placement improvement based on complete re-placement of cell clusters

• US 5,661,663 A
• Filed: 03/24/1995
• Issued: 08/26/1997
• Est. Priority Date: 03/24/1995
• Status: Expired due to Term

• Current Assignee: Bell Semiconductor, LLC (Hilco Inc. (d/b/a Hilco Global))
• Original Assignee: LSI Logic Corporation (Broadcom, Inc.)
• Inventors: Scepanovic, Ranko, Andreev, Alexander E., Koford, James S., Aleshin, Stanislav V., Podkolzin, Alexander S., Kudryavtsev, Valeriy B.
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Fiul, Dan
• Application Number:
  US08/409,757
• Time in Patent Office: 886 Days
• Field of Search: 364/488, 364/489, 364/490, 364/491, 364/419.14, 364/428.28, 395/500
• US Class Current: 716/122
• CPC Class Codes: G06F 30/392 Floor-planning or layout, e.g. partitioning or placement

Abstract

A method of producing a placement of cells for a microelectronic integrated circuit in accordance with specified cell interconnects includes constructing a hierarchial cluster tree based on the interconnects in which a lowest level of the tree includes clusters of interconnected cells, and each successively higher level includes clusters of interconnected clusters from a successively lower level. Clusters in each level are merged by a min-cut operation. Clusters of each successively lower level are then placed within clusters of each successively higher level in progressively decreasing order of levels. Clusters are placed in each level by computing a current gravity point for each cluster in accordance with the cells therein, computing a new gravity point for each cluster in accordance with the current gravity points and the interconnects, and moving each cluster from the current gravity point to the new gravity point. Cells within clusters of the lowest level are then placed by calculating positions of the cells in a row direction, and placing the cells in columns in increasing order of the calculated positions in the row direction. The method is implemented by a parallel processing system including a plurality of parallel processors.

Claims

1. A method of producing a placement of cells for a microelectronic integrated circuit in accordance with specified cell interconnects, comprising the steps:

(a) constructing a hierarchial cluster tree in accordance with said interconnects in which a lowest level of said tree includes clusters of said cells, and each successively higher level includes clusters of clusters from a respective successively lower level;

(b) placing said clusters of each successively lower level within respective clusters of each successively higher level in progressively decreasing order of levels by computing a gravity point for each cluster and modifying cell placement based on a predetermined gravity point adjustment function; and

(c) placing said cells within respective clusters of said lowest level.

2. A method as in claim 1, in which step (a) comprises the substeps of:

(a1) representing said cells and said interconnects as a hypergraph; and

(a2) converting said hypergraph into a graph in which said cells are represented by vertices and said interconnects are represented by edges.

3. A method as in claim 2, wherein said edges are defined in accordance with a predetermined function.

4. A method as in claim 3, wherein said predetermined function comprises defining said edges in a chain configuration.

5. A method as in claim 3, wherein said predetermined function comprises defining said edges in a star configuration.

6. A method as in claim 3, wherein said predetermined function comprises defining said edges in a tree configuration.

7. A method as in claim 1, in which step (a) comprises the substeps of:

(a1) creating disjoint clusters of said cells in accordance with a first predetermined function to produce said lowest level; and

(a2) merging clusters of each successively lower level in accordance with a second predetermined function to produce said clusters of each respective successively higher level.

8. A method as in claim 7, wherein said first predetermined function comprises creating each disjoint cluster as including a selected cell, and cells that are connected to said selected cell by one interconnect respectively.

9. A method as in claim 7, wherein said second predetermined function comprises a min-cut operation.

10. A method as in claim 9, in which step (a2) comprises the substeps, for each cell, of:

(i) computing a first number of intra-cluster interconnects;

(ii) computing a second number of inter-cell interconnects; and

(iii) if said second number is larger than said first number, moving said cell to a cluster to which said cell is connected by a largest number of inter-cell interconnects.

11. A method as in claim 7, in which step (a) further comprises the substep of:

(a3) iteratively repeating step (a2) for each level until said level comprises a number of cells or clusters from lower levels within a predetermined range.

12. A method as in claim 11, in which said predetermined range is approximately 50-100 cells or clusters from lower levels.

13. A method as in claim 7, in which step (a) further comprises the substep of:

(a3) performing step (a2) for progressively higher levels until a highest level is produced which consists of one cluster.

14. A method as in claim 1, wherein step (b) comprises the substeps, for each level of clusters, the steps of:

(b1) generating an initial placement of said clusters; and

(b2) modifying said initial placement in accordance with a predetermined function to produce an improved placement.

15. A method as in claim 1, wherein modifying cell placement based on a predetermined gravity point adjustment function comprises:

(i) computing a new gravity point for each cluster based on the current gravity point and a number of interconnects connecting cluster vertices, wherein clusters are modeled as masses and interconnects are modeled as springs pulling clusters together; and

(ii) moving each cluster from said current gravity point to said new gravity point.

16. A method as in claim 15, in which step (i) further comprises, for each cluster, computing said new gravity point in accordance with directions and distances from said current gravity point to said current gravity points of other clusters to which said cluster is connected, and numbers of interconnects that connect said cluster to said other clusters.

17. A method as in claim 15, further comprising:

(iii) iteratively repeating steps (i) and (ii) until a predetermined criterion is satisfied.

18. A method as in claim 17, in which step (iii) comprises the substeps, performed after each iteration of steps (i) and (ii), of:

(A) computing a quality factor of said placement of said clusters of said level; and

(B) iteratively repeating steps (i) and (ii) if said computed quality factor has a value that is different from a computed quality factor of a previous iteration by more than a predetermined value.

19. A method as in claim 18, wherein said quality factor is a total length of said interconnects of said placement of said clusters of said level.

20. A method as in claim 1, in which step (c) comprises the substeps, for each cluster of said lowest level, of:

(c1) calculating positions of said cells in a first direction; and

(c2) placing said cells along lines which extend in a second direction that is perpendicular to said first direction in increasing order of said calculated positions in said first direction.

21. A physical design automation system for producing a placement of cells for a microelectronic integrated circuit in accordance with specified cell interconnects, comprising:

tree construction means for constructing a hierarchial cluster tree in accordance with said interconnects in which a lowest level of said tree includes clusters of said cells, and each successively higher level includes clusters of clusters from a respective successively lower level;

cluster placement means for placing said clusters of each successively lower cell within respective clusters of each successively higher level in progressively decreasing order of levels by computing a gravity point for each cluster and modifying cell placement based on a predetermined gravity point adjustment function; and

cell placement means for placing said cells respective clusters of said lowest level.

22. A system as in claim 21, in which the tree construction means comprises:

cell representation means for representing said cells and said interconnects as a hypergraph; and

graph conversion means for converting said hypergraph into a graph in which said cells are represented by vertices and said interconnects are represented by edges.

23. A system as in claim 22, wherein said edges are defined in accordance with a predetermined function.

24. A system as in claim 23, wherein said predetermined function comprises means for defining said edges in a chain configuration.

25. A system as in claim 23, wherein said predetermined function comprises means for defining said edges in a star configuration.

26. A system as in claim 23, wherein said predetermined function comprises means for defining said edges in a tree configuration.

27. A system as in claim 21, in which the tree construction means comprises:

pre-clusterization means for creating disjoint clusters of said cells in accordance with a first predetermined function to produce said lowest level; and

cluster modification means for merging clusters of each successively lower level in accordance with a second predetermined function to produce said clusters of each respective successively higher level.

28. A system as in claim 27, in which the first predetermined function comprises means for creating each disjoint cluster as including a selected cell, and cells that are connected to said selected cell by one interconnect respectively.

29. A system as in claim 27, in which the second predetermined function comprises means for merging said clusters using a min-cut operation.

30. A system as in claim 29, in which the cluster modification means comprises:

first computation means for computing, for each cell, a first number of intra-cluster interconnects;

second computation means for computing a second number of inter-cluster interconnects; and

cell movement means for, if said second number is larger than said first number, moving said cell to a cluster to which said cell is connected by a largest number of inter-cell interconnects.

31. A system as in claim 27, in which the tree construction means further comprises:

control means for controlling the cluster modification means to iteratively merge clusters of each level until said level comprises a number of cells or clusters from lower levels within a predetermined range.

32. A system as in claim 31, in which said predetermined range is approximately 50-100 cells or clusters from lower levels.

33. A system as in claim 27, in which the tree construction means further comprises:

control means for controlling the cluster modification means to merge clusters for progressively higher levels until a highest level is produced which consists of one cluster.

34. A system as in claim 21, in which the cluster placement means further comprises:

initial placement generation means for generating an initial placement of said clusters; and

placement modification means for modifying said initial placement in accordance with a predetermined function to produce an improved placement.

35. A system as in claim 21, wherein modifying cell placement based on a predetermined gravity point adjustment function comprises:

new gravity point computation means for computing a new gravity point for each cluster in accordance with said current gravity points and said interconnects based on the current gravity point and a number of interconnects connecting cluster vertices, wherein clusters are modeled as masses and interconnects are modeled as springs pulling clusters together; and

cluster movement means for moving each cluster from said current gravity point to said new gravity point.

36. A system as in claim 35, in which:

the new gravity point computation means comprises means for computing, for each cluster, said new gravity point in accordance with directions and distances from said current gravity point to said current gravity points of other clusters to which said cluster is connected, and numbers of interconnects that connect said cluster to said other clusters.

37. A system as in claim 35, in which the cluster placement means further comprises:

control means for controlling the placement modification means to iteratively modify said initial placement until a predetermined criterion is satisfied.

38. A system as in claim 37, in which the control means comprises:

quality computation means for computing a quality factor of said placement of said clusters of each level; and

iteration means for controlling the placement modification means to iteratively modify said initial placement if said computed quality factor has a value that is different from a computed quality factor of a previous iteration by more than a predetermined value.

39. A system as in claim 38, wherein said quality factor is a total length of said interconnects of said placement of said clusters of said level.

40. A system as in claim 21, in which the cell placement means comprises:

calculation means for calculating positions of said cells in each cluster of said lowest level in a first direction; and

means for placing said cells along lines which extend in a second direction that is perpendicular to said first direction in increasing order of said calculated positions in said first direction.

41. A system as in claim 21, in which the tree construction means, cluster placement means and cell placement means collectively comprise a parallel processing system including a plurality of parallel processors.