Physical design automation system and process for designing integrated circuit chips using fuzzy cell clusterization

• US 5,712,793 A
• Filed: 11/20/1995
• Issued: 01/27/1998
• Est. Priority Date: 11/20/1995
• 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: Kudryvavtsev, Valeriy B., Scepanovic, Ranko, Andreev, Alexander E., Koford, James S., Roseboom, Edward M., Aleshin, Stanislav V., Podkolzin, Alexander S.
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): GARBOWSKI, LEIGH M
• Application Number:
  US08/560,588
• Time in Patent Office: 799 Days
• Field of Search: 364/488, 364/489, 364/490, 364/491
• US Class Current: 716/124
• CPC Class Codes: G06F 30/392 Floor-planning or layout, e.g. partitioning or placement

Abstract

A computer implemented method for generating and placing clusters of cells for integrated circuit design includes providing a netlist including cells, and nets of wires interconnecting the cells. A metric is specified for measuring distance between cells as a function of netlist interconnections. A length of a net is the number of cells interconnected by said net minus one, and a distance between two cells is a sum of lengths of nets that provide a shortest path between the cells. A maximum cluster size criterion, such as maximum distance of a cell from the center of a cluster, is specified to provide a desired amount of overlap between clusters. Clusters of cells are generated, each cluster being generated by designating one of the cells as the center, processing the netlist using the metric to determine distances of cells from the center, and assigning cells having progressively increasing distances from the center to the cluster until the maximum cluster size criterion is reached. Affinities between pairs of clusters are calculated, and the clusters are placed using force directed placement or other methodology based on the affinities.

Claims

1. A computer implemented method for generating a cluster of cells for integrated circuit design, comprising the steps of:

(a) providing a netlist including cells, and nets of wires interconnecting the cells;

(b) designating one of the cells as a center;

(c) specifying a metric for measuring distance between cells as a function of netlist interconnections;

(d) specifying a maximum cluster size criterion;

(e) processing the netlist using said metric to determine distances of cells from the center; and

(f) assigning cells having progressively increasing distances from the center to the cluster until said maximum cluster size criterion is reached.

2. A method as in claim 1, in which:

step (d) comprises specifying said maximum cluster size criterion as a maximum distance from the center;

step (e) comprises processing the netlist using said metric to determine which of the cells are not further than said maximum distance from the center; and

step (f) comprises assigning cells that are not further than said maximum distance from the center to the cluster.

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

(g) defining a length of a net as a number of cells interconnected by said net minus one; and

(h) defining a distance between two cells as a sum of lengths of nets that provide a shortest path between said two cells.

4. A method as in claim 3, in which step (e) comprises the substeps of:

(i) defining a range of a net as a length of said net, plus a shortest distance between any cell in said net and the center;

(j) processing the netlist to determine which of the nets have ranges that do not exceed said maximum distance; and

step (f) comprises assigning cells interconnected by said nets having ranges that are not greater than said maximum distance to the cluster.

5. A method as in claim 4, in which step (j) comprises the substeps of:

(k) determining nets that are connected directly to the center and have ranges less than said maximum distance;

(l) determining nets that are connected directly to said nets determined in step (k) and have ranges less than said maximum distance; and

step (f) comprises assigning cells of said nets determined in steps (k) and (l) to the cluster.

6. A method as in claim 5, in which step (j) further comprises the step of:

(m) determining nets that are connected to the center by progressively increasing numbers of other nets and have ranges less than said maximum distance; and

step (f) comprises assigning cells of said nets determined in step (m) to the cluster.

7. A method as in claim 6, in which steps (k), (l) and (m) comprise defining borders as having indices equal to ranges of nets respectively, and processing nets in said borders having progressively increasing indices.

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

(g) defining a length of a net as a number of cells interconnected by said net minus one; and

(h) defining a distance between two cells as a sum of lengths of nets that provide a shortest path between said two cells.

9. A method as in claim 8, in which step (e) comprises the substeps of:

(i) defining a range of a net as a length of said net, plus a shortest distance between any cell in said net and the center; and

(j) processing the netlist to determine ranges of nets; and

step (f) comprises assigning cells interconnected by said nets having progressively increasing ranges to the cluster until said maximum cluster size criterion is reached.

10. A method as in claim 9, in which step (j) comprises the substeps of:

(k) determining ranges of nets that are connected directly to the center; and

(l) determining ranges of nets that are connected directly to said nets determined in step (k); and

step (f) comprises assigning cells of said nets determined in steps (k) and (l) to the cluster until said maximum cluster size criterion is reached.

11. A method as in claim 10, in which step (j) further comprises the step of:

(m) determining nets that are connected to the center by progressively increasing numbers of other nets; and

step (f) comprises assigning cells of said nets determined in step (m) to the cluster until said maximum cluster size criterion is reached.

12. A method as in claim 11, in which steps (k), (l) and (m) comprise defining borders as having indices equal to ranges of nets respectively, and processing nets in said borders having progressively increasing indices.

13. A method as in claim 1, in which:

step (d) comprises specifying said maximum cluster size criterion as a total maximum area of cells assigned to the cluster; and

step (f) comprises assigning cells to the cluster until a total area of said cells assigned to the cluster is not smaller than said total maximum area.

14. A method as in claim 1, in which:

step (d) comprises specifying said maximum cluster size criterion as a maximum total height of cells assigned to the cluster; and

step (f) comprises assigning cells to the cluster until a total height of said cells assigned to the cluster is not smaller than said total maximum height.

15. A method as in claim 1, in which:

step (d) comprises specifying said maximum cluster size criterion as a maximum number of cells assigned to the cluster; and

step (f) comprises assigning cells to the cluster until a number of said cells assigned to the cluster is not smaller than said maximum number.

16. A computer implemented method for generating clusters of cells for integrated circuit design, comprising the steps of:

(a) providing a netlist including cells, and nets of wires interconnecting the cells;

(b) specifying a metric for measuring distance between cells as a function of netlist interconnections;

(c) specifying a maximum cluster size criterion which provides a desired amount of overlap between clusters; and

(d) generating clusters of cells, each cluster being generated by performing the steps of;

(e) designating one of the cells as a center;

(f) processing the netlist using said metric to determine distances of cells from the center; and

(g) assigning cells having progressively increasing distances from the center to the cluster until said maximum cluster size criterion is reached.

17. A method as in claim 16, in which step (e) comprises designating a cell for a new cluster as being further from previously generated clusters than a predetermined distance.

18. A method as in claim 16, in which steps (c) and (d) comprise generating said clusters such that said amount of overlap is zero.

19. A method as in claim 16, in which:

step (c) comprises specifying said maximum cluster size criterion as a maximum distance from the center;

step (e) comprises processing the netlist using said metric to determine which of the cells are not further than said maximum distance from the center; and

step (f) comprises assigning cells that are not further than said maximum distance from the center to the cluster.

20. A method as in claim 19, in which step (b) comprises the substeps of:

(h) defining a length of a net as a number of cells interconnected by said net minus one; and

(i) defining a distance between two cells as a sum of lengths of nets that provide a shortest path between said two cells.

21. A method as in claim 20, in which step (f) comprises the substeps of:

(j) defining a range of a net as a length of said net, plus a shortest distance between any cell in said net and the center;

(k) processing the netlist to determine which of the nets have ranges that do not exceed said maximum distance; and

step (g) comprises assigning cells interconnected by said nets having ranges that are not greater than said maximum distance to the cluster.

22. A method as in claim 21, in which step (k) comprises the substeps of:

(l) determining nets that are connected directly to the center and have ranges less than said maximum distance;

(m) determining nets that are connected directly to said nets determined in step (k) and have ranges less than said maximum distance; and

step (g) comprises assigning cells of said nets determined in steps (l) and (m) to the cluster.

23. A method as in claim 22, in which step (k) further comprises the step of:

(n) determining nets that are connected to the center by progressively increasing numbers of other nets and have ranges less than said maximum distance; and

step (g) comprises assigning cells of said nets determined in step (m) to the cluster.

24. A method as in claim 23, in which steps (l), (m) and (n) comprise defining borders as having indices equal to ranges of nets respectively, and processing nets in said borders having progressively increasing indices.

25. A method as in claim 16, in which step (b) comprises the substeps of:

(h) defining a length of a net as a number of cells interconnected by said net minus one; and

(i) defining a distance between two cells as a sum of lengths of nets that provide a shortest path between said two cells.

26. A method as in claim 25, in which step (f) comprises the substeps of:

(j) defining a range of a net as a length of said net, plus a shortest distance between any cell in said net and the center; and

(k) processing the netlist to determine ranges of nets; and

step (g) comprises assigning cells interconnected by said nets having progressively increasing ranges to the cluster until said maximum cluster size criterion is reached.

27. A method as in claim 26, in which step (k) comprises the substeps of:

(l) determining ranges of nets that are connected directly to the center; and

(m) determining ranges of nets that are connected directly to said nets determined in step (k); and

step (g) comprises assigning cells of said nets determined in steps (k) and (l) to the cluster until said maximum cluster size criterion is reached.

28. A method as in claim 27, in which step (k) further comprises the step of:

(n) determining nets that are connected to the center by progressively increasing numbers of other nets; and

step (g) comprises assigning cells of said nets determined in step (m) to the cluster until said maximum cluster size criterion is reached.

29. A method as in claim 28, in which steps (l), (m) and (n) comprise defining borders as having indices equal to ranges of nets respectively, and processing nets in said borders having progressively increasing indices.

30. A method as in claim 16, in which:

step (c) comprises specifying said maximum cluster size criterion as a total maximum area of cells assigned to the cluster; and

step (g) comprises assigning cells to the cluster until a total area of said cells assigned to the cluster is not smaller than said total maximum area.

31. A method as in claim 16, in which:

step (c) comprises specifying said maximum cluster size criterion as a maximum total height of cells assigned to the cluster; and

step (g) comprises assigning cells to the cluster until a total height of said cells assigned to the cluster is not smaller than said total maximum height.

32. A method as in claim 16, in which:

step (c) comprises specifying said maximum cluster size criterion as a maximum number of cells assigned to the cluster; and

step (g) comprises assigning cells to the cluster until a number of said cells assigned to the cluster is not smaller than said maximum number.

33. A method as in claim 16, further comprising the steps of:

(h) calculating affinities between each pair of clusters; and

(i) determining a placement of said clusters such that a cost function based on said affinities is substantially minimized.

34. A method as in claim 33, in which step (h) comprises calculating an affinity between two clusters as a number of cells which are common to said two clusters divided by a function of the numbers of cells in said two clusters.

35. A method as in claim 34, in which said function is a square root of a product of said numbers of cells in said two clusters.

36. A method as in claim 33, in which step (i) comprises determining said placement of said clusters using a force directed placement methodology.