Joseph Gebis

EECS Department, University of California, Berkeley

Technical Report No. UCB/CSD-03-1288

2003

This publication is archived. It is kept only for reference purposes, so it is no longer being updated and may not meet accessibility standards. If you need this content in a different format, please email webteam@eecs.berkeley.edu.

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/Archive/CSD-03-1288.pdf

VIRAM1 (Vector Intelligent RAM 1) is a low-power multimedia processor with embedded DRAM designed at UC Berkeley in 2002 and fabricated in 2003. It includes a scalar core and four vector computation units, called lanes. The goals of the chip, low-power media processing, require that the vector lanes have efficient integer multipliers that can work with a variety of data sizes. In this report, I describe an efficient partitionable integer multiplier that is designed to work in VIRAM1's vector computation lanes. The multiplier is capable of operating with a latency of two cycles at 200 MHz in a 1.2 V, .18 micron process at 64, 32, or 16 bit data width sizes, while consuming less than 250 mW of power. I describe and evaluate design options for different parts of the multiplier, and analyze the results of the chosen options.


BibTeX citation:

@techreport{Gebis:CSD-03-1288,
    Author= {Gebis, Joseph},
    Title= {A 64-bit Partitionable Integer Multiplier for VIRAM1},
    Year= {2003},
    Month= {Nov},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/5698.html},
    Number= {UCB/CSD-03-1288},
    Abstract= {VIRAM1 (Vector Intelligent RAM 1) is a low-power multimedia processor with embedded DRAM designed at UC Berkeley in 2002 and fabricated in 2003. It includes a scalar core and four vector computation units, called lanes. The goals of the chip, low-power media processing, require that the vector lanes have efficient integer multipliers that can work with a variety of data sizes. In this report, I describe an efficient partitionable integer multiplier that is designed to work in VIRAM1's vector computation lanes. The multiplier is capable of operating with a latency of two cycles at 200 MHz in a 1.2 V, .18 micron process at 64, 32, or 16 bit data width sizes, while consuming less than 250 mW of power. I describe and evaluate design options for different parts of the multiplier, and analyze the results of the chosen options.},
}

EndNote citation:

%0 Report
%A Gebis, Joseph 
%T A 64-bit Partitionable Integer Multiplier for VIRAM1
%I EECS Department, University of California, Berkeley
%D 2003
%@ UCB/CSD-03-1288
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/5698.html
%F Gebis:CSD-03-1288