Recursive Bisection Based Mixed Block Placement Ateen Khatkhate 1 Chen Li 2 Ameya R. Agnihotri 1 Mehmet C. Yildiz 3 Satoshi Ono 1 Cheng-Kok Koh 2 Patrick H. Madden 1 SUNY Binghamton CSD 1 Purdue University ECE 2 IBM Austin Research Lab 3 ABSTRACT Many current designs contain a large number of standard cells in- termixed with larger macro blocks. The range of size in these “mixed block” designs complicates the placement process consid- erably; traditional methods produce results that are far from satis- factory. In this paper we extend the traditional recursive bisection stan- dard cell placement tool Feng Shui to directly consider mixed block designs. On a set of recent benchmarks, the new version obtains placements with wire lengths substantially lower than other current tools. Compared to Feng Shui 2.4, the placements of a Capo-based approach have 29% higher wire lengths, while the placements of mPG are 26% higher. Run times of our tool are also lower, and the general approach is scalable. Categories and Subject Descriptors J.6 [Computer-Aided Engineering]: CAD General Terms Algorithms Keywords Placement, floorplanning, mixed block design 1. INTRODUCTION There has been explosive growth in the size of integrated circuits; following the exponential curve of Moore’s law, modern designs can have over 50 million transistors. This growth is projected to continue, and circuit designers are having difficulty using this ca- pacity effectively. Heirarchy is commonly used to tame unwieldly designs. At the highest heirarchical level, floorplanning of a few hundred large blocks can be done effectively. At the lowest level, place- ment tools can handle hundreds of thousands or millions of stan- dard cells. Between these two extremes is mixed block placement, Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. ISPD’04 April 18–21,2004,Phoenix,Arizona,USA. Copyright 2004 ACM 1-58113-817-2/04/0004 ...$5.00. in which blocks of moderate size are intermixed with many stan- dard cells. Current placement tools have had difficulty in handling this “boulders and dust” problem. In this paper, we adapt an earlier version of our recursive bi- section based standard cell placement tool Feng Shui to handle mixed block designs. While many recent academic approaches have sought to address standard cell and macro block placement in separate steps, we consider them simultaneously. We find that this can be done more easily than might be expected, and produces results that are vastly superior to recently published work. As our approach is based on recursive bisection, the tool itself is quite fast and scales well to large designs. Using benchmarks derived from industry partitioning examples, our new placement tool obtains excellent wire length results. A multi-stage flow using the standard cell placement tool Capo, and an integrated approach called mPG, obtain results that are respec- tively 29% and 26% higher on average. On some benchmarks, wire lengths obtained by our tool are roughly half of those of Capo and mPG. The gap in solution quality may be surprising; in most areas of design automation, we might expect only a few percent improve- ment from a new technique. We have verified our results with pub- licly available tools, and our placement results and the placement tool itself are available through the web. We suspect that the magnitude of improvement is largely due to the limited amount of recent published research in the area: most academic groups have focused on either standard cell placement or block placement, with mixed block problems getting compara- tively little attention. Inspection of our placements reveals many opportunities for further gains, and we anticipate that there will be additional significant improvement in mixed block placement. The remainder of this paper is organized as follows. In Section 2, we describe various types of placement problems, and prior work related to mixed block designs. Section 3 describes our approach to mixed block placement. Experimental results are given in Section 4, and we conclude the paper with Section 5. 2. PREVIOUS WORK Circuit placement is a well studied problem, and comes in a va- riety of forms. Most common are standard cell placement, block placement, and our focus here, mixed block placement. 2.1 Standard Cell Placement In standard cell placement, we may have a great many cells, small rectangular blocks that are of uniform height, but possibly varying width. Each cell contains the circuitry for a relatively sim- ple logical function, and the cells are packed into rows much as one might use bricks to form a wall. The desired circuit functionality is obtained by connecting each cell with metal wiring. The arrange-