\documentclass[twocolumn]{article} 
% \documentclass{article}

\setlength{\oddsidemargin}{-0.5in}
\setlength{\evensidemargin}{-0.5in}
\setlength{\topmargin}{0.0in}
\setlength{\headheight}{0in}
\setlength{\headsep}{0in}
\setlength{\textwidth}{7.0in}
\setlength{\textheight}{9.5in}
\setlength{\parindent}{0in}
\setlength{\parskip}{0.05in}
\setlength{\columnseprule}{0.3pt}

\setlength\columnsep{25pt}

\usepackage{fancyvrb}
\usepackage{relsize}
\usepackage{hyperref}

\usepackage{listings}
\usepackage{amsmath}

\begin{document}

\lstset{xleftmargin=5mm,framexleftmargin=10mm,basicstyle=\footnotesize}

Name: \_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_

Directions: MAKE SURE TO COPY YOUR ANSWERS TO A SEPARATE SHEET FOR SENDING
ME AN ELECTRONIC COPY LATER.

{\bf 1.} (20) The NVIDIA GPU execution model is Single Instruction
Multiple Threads, SIMT.  What is the acronym for MPI?

{\bf 2.} (20) The process ID in MPI has a 2-level hierarchical structure,
rather analogous to {\bf blockIdx} and {\bf threadIdx} in CUDA.  What
are the official MPI terms for the two levels? (Answer on one line.)

{\bf 3.} (20) Consider the run-length decoding examples in Secs.\ 10.5 and
10.6, the first in OMP and the second in Thrust.  State the line number
in Sec.\ 10.6 that is analogous to line 10 in Sec.\ 10.5.  If there is no
such line in 10.6, answer None.

{\bf 4.} (20) In the example titled, Removing 0s from an Array, Sec.\
8.4, consider what may happen in terms of the order of the nonzero
values within the {\bf no0s} array.  Which one of the following is true?

\begin{itemize}

\item [(i)] The order will be the same as in the {\bf has0s} array.  For
instance, if 12 and 5 were in {\bf has0s}, with 12 in a lower-index
position than 5, then the index of 12 in {\bf no0s} will be lower than
that of 5.

\item [(ii)] The order will be reversed.

\item [(iii)] The order will be random, in the sense that in several runs
of the program, different orders may occur in different runs.

\item [(iv)] The order will be nonrandom, but neither (i) nor (ii) will
necessarily occur.

\end{itemize}

{\bf 5.} (20) In this problem you will develop an OpenMP function that
works like {\bf thrust::copy\_if()}.  Fill in the blanks:

\begin{lstlisting}
#include <stdio.h>  
#include <omp.h>  

int u[8] = {1,0,1,1,0,1,0,1},
    v[8] = {1,5,6,3,0,2,0,8},  
    w[8];

int f(int a) {return a != 0;}

void omp_copy_if(int *x1, int *x2, int nin,
                 int *y, int *nout,
                 int (*boolftn)(int)) {
   blank (a)
   #pragma omp parallel 
   {  
      int i;
      #pragma omp for
      for (i = 0; i < nin; i++)
         if (  blank (b)
            blank (c)
            ntrue++;
            blank (d)
         }
      #pragma omp single
      *nout = ntrue;
   }
}

main(int argc, char **argv)
{  int i,no;
   omp_copy_if(v,u,8,w,&no,f);
   // should print 1 6 3 2 8 
   for (i = 0; i < no; i++) printf("%d\n",w[i]);
}  

}}
\end{lstlisting}

\onecolumn

\newpage

{\bf Solutions:}

{\bf 1.} SPMD

{\bf 2.} communicator, rank

{\bf 3.} None; Thrust does not work directly with thread number.

{\bf 4.} (i)

{\bf 5.

\begin{lstlisting}

#include <stdio.h>  
#include <omp.h>  

int u[8] = {1,0,1,1,0,1,0,1},
    v[8] = {1,5,6,3,0,2,0,8},  
    w[8];

int f(int a) {return a != 0;}

void omp_copy_if(int *x1, int *x2, int nin,
                 int *y, int *nout,
                 int (*bool\footnote{)(int)) {
   int ntrue = 0;
   #pragma omp parallel 
   {  
      int i;
      #pragma omp for
      for (i = 0; i < nin; i++)
         if (boolftn(x2[i])) {  
            #pragma omp atomic
            ntrue++;
            y[ntrue-1] = x1[i];  
         }
      #pragma omp single
      *nout = ntrue;
   }
}

main(int argc, char **argv)
{  int i,no;
   omp_copy_if(v,u,8,w,&no,f);
   // should print 1 6 3 2 8 
   for (i = 0; i < no; i++) printf("%d\n",w[i]);
}  

}}

\end{lstlisting}}

\end{document}