\relax 
\citation{beowulf}
\@writefile{toc}{\contentsline {section}{\numberline {1}Introduction}{1}}
\@writefile{toc}{\contentsline {section}{\numberline {2}Amdahl's Law \& Parallel Speedup}{2}}
\citation{Amdahl}
\newlabel{rate}{{3}{3}}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces $T_{is} = 0$ and $T_p =$ 10, 100, 1000, 10000, 100000 (in increasing order).}}{4}}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces $T_{is} = 10$ and $T_p =$ 10, 100, 1000, 10000, 100000 (in increasing order).}}{4}}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces $T_{is} = 10$ and $T_p =$ 10, 100, 1000, 10000, 100000 (in increasing order) with $T_{is}$ contributing {\em  quadratically} in $N$.}}{5}}
\@writefile{toc}{\contentsline {section}{\numberline {3}Microbenchmarking Tools}{5}}
\citation{ATLAS}
\citation{lmbench}
\citation{netperf}
\citation{cpu-rate}
\citation{Eden}
\citation{Eden}
\citation{beowulf}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.1}Lmbench Results}{7}}
\@writefile{lot}{\contentsline {table}{\numberline {1}{\ignorespaces Lucifer System Description}}{7}}
\@writefile{lot}{\contentsline {table}{\numberline {2}{\ignorespaces lmbench latencies for selected processor/process activities. The values are all times in microseconds averaged over ten independent runs (with error estimates provided by an unbiased standard deviation), so ``smaller is better''.}}{7}}
\@writefile{lot}{\contentsline {table}{\numberline {3}{\ignorespaces Lmbench latencies for context switches, in microseconds (smaller is better).}}{7}}
\@writefile{lot}{\contentsline {table}{\numberline {4}{\ignorespaces Lmbench {\em  local} communication latencies, in microseconds (smaller is better).}}{7}}
\@writefile{lot}{\contentsline {table}{\numberline {5}{\ignorespaces Lmbench {\em  network} communication latencies, in microseconds (smaller is better).}}{8}}
\@writefile{lot}{\contentsline {table}{\numberline {6}{\ignorespaces Lmbench {\em  memory} latencies in nanoseconds (smaller is better). Also see graphs for more complete picture.}}{8}}
\@writefile{lot}{\contentsline {table}{\numberline {7}{\ignorespaces Lmbench {\em  local} communication bandwidths, in $10^6$ bytes/second (bigger is better).}}{8}}
\@writefile{lot}{\contentsline {table}{\numberline {8}{\ignorespaces Lmbench {\em  network} communication bandwidths, in $10^6$ bytes/second (bigger is better).}}{8}}
\@writefile{lot}{\contentsline {table}{\numberline {9}{\ignorespaces CPU-rates in BOGOMFLOPS -- $10^6$ simple arithmetic operations/second, in L1 cache (bigger is better). Also see graph for out-of-cache performance.}}{8}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.2}Netperf Results}{8}}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces TCP Stream (netperf) measurements of bandwidth as a function of packet size between lucifer and eve.}}{9}}
\newlabel{tcp_stream}{{4}{9}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.3}CPU Results}{9}}
\citation{cpu-rate}
\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces Double precision floating point operations per second as a function of vector length (in bytes). All points average 100 independent runs. The dashed lines indicate the locations of the L1 and L2 cache boundaries.}}{10}}
\newlabel{mdops}{{5}{10}}
\@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces The standard deviation (error) associated with figure 5\hbox {}.}}{10}}
\newlabel{mdops_sigma}{{6}{10}}
\citation{ATLAS}
\citation{profiling}
\@writefile{toc}{\contentsline {section}{\numberline {4}Conclusions}{11}}
\bibcite{beowulf}{beowulf}
\bibcite{Amdahl}{Amdahl}
\bibcite{Amalsi}{Amalsi}
\bibcite{Foster}{Foster}
\bibcite{Kumar}{Kumar}
\bibcite{lmbench}{lmbench}
\bibcite{netperf}{netperf}
\bibcite{cpu-rate}{cpu-rate}
\bibcite{Eden}{Eden}
\bibcite{profiling}{profiling}
\bibcite{ATLAS}{ATLAS}
\@writefile{toc}{\contentsline {section}{\numberline {5}Acknowledgments}{12}}
\@writefile{toc}{\contentsline {section}{\numberline {6}Availability}{12}}