2003Unpublished venueRequires access

Using partitioning to help convergence in the standard-cell design automation methodology

H. Kapadia, Mark Horowitz

Open publisher page 6 citations

Abstract

This paper explores a standard-cell design methodology based on netlist partitioning as a solution for the problem of lack of convergence in the conventional methodology in deep submicron technologies. A synthesized design block is partitioned along unpredictable nets that are identified from the netlist structure. The size of each partition is restricted so that the longest possible local net in a partition can be sufficiently driven by an average library gate, hence allowing statistical wire-load modeling for the local nets. The block is resynthesized using a hybrid wire-load model that takes into account accurate wire-load information on the unpredictable nets derived after floorplanning the partitions, and uses custom statistical wire-load models within each partition. Final placement is restricted to respect the initial floorplan. The methodology was implemented using existing commercial tools for synthesis and layout. Experimental results show high correlation between synthesis estimates and post-placement measurements of wire-loads and gate delays with the new methodology. The trade-offs of partitioning, current limitations of the methodology and future work to overcome these limitations are also discussed.

About this research paper

What this paper is about

This paper explores a standard-cell design methodology based on netlist partitioning as a solution for the problem of lack of convergence in the conventional methodology in deep submicron technologies. A synthesized design block is partitioned along unpredictable nets that are identified from the netlist structure. The size of each partition is restricted so that the longest possible local net in a partition can be sufficiently driven by an average library gate, hence allowing statistical wire-load modeling for the local nets. The block is resynthesized using a hybrid wire-load model that takes into account accurate wire-load information on the unpredictable nets derived after floorplanning the partitions, and uses custom statistical wire-load models within each partition. Final placement is restricted to respect the initial floorplan. The methodology was implemented using existing commercial tools for synthesis and layout. Experimental results show high correlation between synthesis estimates and post-placement measurements of wire-loads and gate delays with the new methodology. The trade-offs of partitioning, current limitations of the methodology and future work to overcome these limitations are also discussed.

Why it matters

OpenAlex reports 6 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

This paper explores a standard-cell design methodology based on netlist partitioning as a solution for the problem of lack of convergence in the conventional methodology in deep submicron technologies. A synthesized design block is partitioned along unpredictable nets that are identified from the netlist structure. The size of each partition is restricted so that the longest possible local net in a partition can be sufficiently driven by an average library gate, hence allowing statistical wire-load modeling for the local nets. The block is resynthesized using a hybrid wire-load model that takes into account accurate wire-load information on the unpredictable nets derived after floorplanning the partitions, and uses custom statistical wire-load models within each partition. Final placement is restricted to respect the initial floorplan. The methodology was implemented using existing commercial tools for synthesis and layout. Experimental results show high correlation between synthesis estimates and post-placement measurements of wire-loads and gate delays with the new methodology. The trade-offs of partitioning, current limitations of the methodology and future work to overcome these limitations are also discussed.

Key concepts: Netlist, Floorplan, Partition (number theory), Standard cell, Computer science, Electronic design automation, Integrated circuit layout, Block (permutation group theory)

Related papers

Back to paper searchBrowse research topicsOriginal source
Using partitioning to help convergence in the standard-cell design automation methodology — Research Paper | ScholarLens