Ignore:
Timestamp:
Jan 24, 2014, 11:07:08 AM (6 years ago)
Author:
cameron
Message:

Some cleanups

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/Working/re/avx2.tex

    r3515 r3617  
    1919\paragraph*{AVX2 Stream Addition}
    2020 \begin{figure*}[tbh]
    21 \begin{center}
     21
     22\begin{center} \small
    2223\begin{verbatim}
    2324void add_ci_co(bitblock_t x, bitblock_t y, carry_t carry_in, carry_t & carry_out, bitblock_t & sum) {
     
    3637  sum = simd256<64>::add(partial_sum, _mm256_cvtepu16_epi64(avx_select_lo128(convert(spread))));
    3738}
    38 
    3939\end{verbatim}
    40 
    4140\end{center}
    4241\caption{AVX2 256-bit Addition}
     
    6564tick label style={font=\tiny},
    6665enlarge x limits=0.15,
    67 enlarge y limits={0.15, upper},
     66%enlarge y limits={0.15, upper},
    6867ymin=0,
    6968legend style={at={(0.5,-0.15)},
     
    116115tick label style={font=\tiny},
    117116enlarge x limits=0.15,
    118 enlarge y limits={0.15, upper},
     117%enlarge y limits={0.15, upper},
    119118ymin=0,
    120119legend style={at={(0.5,-0.15)},
     
    140139instruction count was reflected in a considerable speed-up
    141140in the bitstreams implementation.  However, the speed-up was
    142 considerably less than expected.  As shown in \label{fig:AVXIPC}
     141considerably less than expected.  As shown in \ref{fig:AVXIPC}
    143142the AVX2 version has lost some of the superscalar efficiency
    144143of the SSE2 code.   This is a performance debugging issue
     
    155154tick label style={font=\tiny},
    156155enlarge x limits=0.15,
    157 enlarge y limits={0.15, upper},
     156%enlarge y limits={0.15, upper},
    158157ymin=0,
    159158legend style={at={(0.5,-0.15)},
Note: See TracChangeset for help on using the changeset viewer.