Optimization processing for integrated circuit physical design automation system using chaotic fitness improvement method

• US 5,682,322 A
• Filed: 04/19/1994
• Issued: 10/28/1997
• Est. Priority Date: 04/19/1994
• 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, Koford, James S., Boyle, Douglas B., Rostoker, Michael D., Jones, Edwin R.
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): GARBOWSKI, LEIGH M
• Application Number:
  US08/229,949
• Time in Patent Office: 1,288 Days
• Field of Search: 364/488-491
• US Class Current: 716/123
• CPC Class Codes: G06F 30/392 Floor-planning or layout, e.g. partitioning or placement

Abstract

The fitness of a cell placement for an integrated circuit chip is optimized by relocating at least some of cells to new locations that provide lower interconnect congestion. For each cell, the centroid of the net of cells to which the cell is connected is computed. The cell is then moved toward the centroid by a distance that is equal to the distance from the current position of the cell to the centroid multiplied by a "chaos" factor λ. The value of λ is selected such that the cell relocation operations will cause the placement to converge toward an optimal configuration without chaotic diversion, but with a sufficiently high chaotic element to prevent the optimization operation from becoming stuck at local fitness maxima. The new cell locations can be modified to include the effects of cells in other locations, such as by incorporating a function of cell density gradient or force direction into the computation. This spreads out clumps of cells so that the density of cells is more uniform throughout the placement. The attraction between cells in the nets is balanced against repulsion caused by a high local cell density, providing an optimized tradeoff of wirelength, feasibility and congestion.

Claims

1. A method of relocating a cell in a cell placement for an integrated circuit chip, comprising the step of:

(a) computing a centroid as a first predetermined function of relationships between said cell and other associated cells on the chip;

(b) computing a first length from a current location of said cell to said centroid;

(c) computing a second length as a second predetermined function of said first length; and

(d) moving said cell from said current location toward said centroid by said second length.

2. A method as in claim 1, in which step (a) comprises computing said centroid as a center of gravity of said associated cells.

3. A method as in claim 1, in which step (c) comprises computing said second length as being proportional to said first length.

4. A method as in claim 1, in which step (c) comprises computing said second length as being between approximately 0.5 to 1.5 times said first length.

5. A method as in claim 1, further comprising the steps of:

(e) computing a density gradient of cells in locations that are proximate to said cell, said density gradient having a magnitude and a decreasing density direction; and

(f) moving said cell from said current position in said decreasing density direction by a third distance that is a third predetermined function of said magnitude.

6. A method as in claim 5, in which said magnitude for each location in step (e) is equal to a number of cells in said location respectively.

7. A method as in claim 5, in which said third length is proportional to said magnitude.

8. A method as in claim 1, further comprising the steps of:

(e) computing a net force on said cell as a third predetermined function of lengths from said cell to cells in locations that are proximate to said cell, said net force having a magnitude and a direction; and

(f) moving said cell from said current position in said direction by a third length that is a fourth predetermined function of said magnitude.

9. A method as in claim 8, in which said third length is proportional to said magnitude.

10. A method as in claim 8, in which said magnitude decreases as said lengths from said cell to said cells in locations that are proximate to said cell increase.

11. A method as in claim 1, further comprising the steps of:

(e) computing a net force on said cell as a third predetermined function of distances from said cell to all other cells of said placement, said net force having a magnitude and a direction; and

(f) moving said cell from said current position in said direction by a third length that is a fourth predetermined function of said magnitude.

12. A method as in claim 11, in which said third length distance is proportional to said magnitude.

13. A method as in claim 11, in which said magnitude decreases as said distances from said cell to said all other cells of said placement increase.

14. A method as in claim 1, further comprising the step of repeating steps (a) through (d) until a predetermined criteria is achieved.

15. A method as in claim 1 wherein said second predetermined function comprises a chaotic function preventing entrapment of the method at a local fitness optima.

16. A method of increasing the fitness of a cell placement for an integrated circuit chip, comprising the steps of:

(a) computing centroids for a plurality of cells in said placement as a first predetermined function of relationships between said cells and other associated cells on the chip;

(b) computing first lengths from current locations of said cells to said centroids respectively;

(c) computing second lengths as a second predetermined function of said first lengths respectively; and

(d) moving said cells from said current locations toward said centroids by said second lengths respectively.

17. A method as in claim 19, in which step (e) comprises distributing said cells into locations that are substantially closest to locations that said cells occupied after performing step (d).

18. A method as in claim 16, in which step (a) comprises selecting said plurality of cells randomly.

19. A method as in claim 16, in which step (a) comprises selecting said plurality in accordance with a predetermined fitness criterion.

20. A method as in claim 16, in which step (a) comprises selecting said plurality of cells in a predetermined order.

21. A method as in claim 16, in which step (a) comprises selecting said plurality of cells in accordance with said locations of cells to which said cells are connected respectively.

22. A method as in claim 16, in which steps (a) to (d) are performed sequentially on said plurality of cells.

23. A method as in claim 16, in which steps (a) to (d) are performed in parallel on said plurality of cells.

24. A method as in claim 16, further comprising the steps, performed after step (d), of:

(e) selecting a new plurality of cells; and

(f) repeating steps (a) to (d) for said new plurality of cells.

25. A method as in claim 24, further comprising the step, performed after step (f), of:

(g) repeating steps (e) and (f) a predetermined number of times.

26. A method as in claim 24, further comprising the steps, performed after step (f), of:

(g) measuring said fitness of said placement; and

(h) repeating steps (e) and (f) if said fitness is below a predetermined value.

27. A method as in claim 26, in which step (g) comprises measuring said fitness as a predetermined function of cell interconnect congestion.

28. A method as in claim 24, further comprising the steps, performed after step (f), of:

(g) measuring said fitness of said placement; and

(h) repeating steps (e) and (f) if said fitness has changed by more than a predetermined value.

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

steps (a) and (d) allow more than one cell to be moved to a single location, and allow locations to have no cells thereat; and

the method further comprises the step, performed after step (d) of;

(e) distributing all cells of said placement into said locations respectively.

30. A method of relocating an entity in a permutation of a predetermined number of entities that are arranged in a pattern of locations respectively, comprising the steps of:

(a) computing a centroid as a first predetermined function of relationships between said entity and other associated entities in a predetermined manner;

(b) computing a first length from a current location of said entity to said centroid;

(c) computing a second length as a second predetermined function of said first length; and

(d) moving said entity from said current location toward said centroid by said second length.

31. A method as in claim 30, in which step (a) comprises computing said centroid as a center of gravity of said associated entities.

32. A method as in claim 30, in which step (c) comprises computing said second length as being proportional to said first distance.

33. A method as in claim 30, further comprising the step of repeating steps (a) through (d) until a predetermined criteria is achieved.

34. A method as in claim 30 wherein said second predetermined function comprises a chaotic function preventing entrapment of the method at a local fitness optima.