Changeset 1395 for docs/HPCA2012


Ignore:
Timestamp:
Aug 30, 2011, 4:14:00 PM (8 years ago)
Author:
ashriram
Message:

Changed intro 1st line

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/HPCA2012/01-intro.tex

    r1393 r1395  
    11\section{Introduction}
    2 We have now long since reached the limit to classical Dennard voltage
    3 scaling that enabled us to keep all of transistors afforded by Moore's
    4 law active. This has already resulted in a rethink of the way
    5 general-purpose processors are built: processor frequencies have
    6 remained stagnant over the last 5 years with the capability to boost
    7 a core's frequency only if other cores on the chip are shut off. Chip makers
    8 strive to achieve energy efficient computing by operating at more
    9 optimal core frequencies and aim to increase performance with a larger
    10 number of cores. Unfortunately, given the limited levels of
    11 parallelism that can be found in applications~\cite{blake-isca-2010},
    12 it is not certain how many cores can be productively used in scaling
    13 our chips~\cite{esmaeilzadeh-isca-2011}. This is because exploiting
     2Classical Dennard voltage scaling enabled us to keep all of transistors
     3afforded by Moore's law active. Dennard scaling reached its limits in
     42005 and this has resulted in a rethink of the way general-purpose
     5processors are built: frequencies have remained stagnant
     6over the last 5 years with the capability to boost a core's frequency
     7only if other cores on the chip are shut off. Chip makers strive to
     8achieve energy efficient computing by operating at more optimal core
     9frequencies and aim to increase performance with a larger number of
     10cores. Unfortunately, given the limited levels of parallelism that can
     11be found in applications~\cite{blake-isca-2010}, it is not certain how
     12many cores can be productively used in scaling our
     13chips~\cite{esmaeilzadeh-isca-2011}. This is because exploiting
    1414parallelism across multiple cores tends to require heavy weight
    1515threads that are difficult to manage and synchronize.
Note: See TracChangeset for help on using the changeset viewer.