Ignore:
Timestamp:
Feb 24, 2014, 2:39:02 PM (5 years ago)
Author:
cameron
Message:

mCleannups

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/Working/re/re-main.tex

    r3645 r3649  
    277277We also have adapted our long-stream addition technique
    278278to perform 4096-bit additions using 64 threads working in lock-step
    279 SIMT fashion.  A similar technique is known to the GPGPU programming
    280 community\cite{}.   
     279SIMT fashion. 
    281280 
    282281\begin{figure}[tbh]
     
    446445
    447446We use the following general model using SIMD methods for constant-time
    448 long-stream addition up to 4096 bits.   Related solutions have been
    449 independently developed on GPUs
    450 (\verb`http://stackoverflow.com/questions/12957116/` verb`-integer-addition-with-cuda`),
    451 however our model is intended to be a more broady applicable abstraction.
     447long-stream addition up to 4096 bits.   Related GPGPU solutions have been
     448independently developed\cite{Crovella2012},
     449however our model is intended to be a more broadly applicable abstraction.
    452450We assume the availability of the following SIMD/SIMT operations
    453451operating on vectors of $f$ 64-bit fields.
     
    477475the parallel units.   There are a variety of ways in which
    478476these facilities may be implemented depending on the
    479 underlying architecture; details of our AVX2 and GPU implementations
     477underlying architecture; details of our AVX2 and GPGPU implementations
    480478are presented later.   
    481479
     
    569567expression matching as shown herein, it seems reasonable to expect
    570568such instructions to become available.    Alternatively, it may
    571 be worthwhile to simply ensure that the \verb#hsimd<64>::mask(X)#
     569be worthwhile to simply ensure that the \verb#hsimd<64>::mask(X)# and
    572570\verb#simd<64>::spread(X)# operations are efficiently supported.
    573571
     
    653651file {data/gputime.dat};
    654652
    655 \legend{SSE2,AVX2,GPU,Annot}
     653\legend{SSE2,AVX2,GPGPU,Annot}
    656654\end{axis}
    657655\end{tikzpicture}
Note: See TracChangeset for help on using the changeset viewer.