You need to re$-$write the assembly code, which you build in the previous exercise, using the RISC$-$V V

extension. Before you write the code, you are suggested to study the RISC$-$V V extension document $($from p$.$

$10$ to p$.31$ and p$.55)$ to get familiar with the concept of RISC$-$V vector programming. After you write thevectorized program, you should collect the performance data and derive related performance statistics as you

did in the previous exercise.

As shown in the improved$\_$version$1$ function, you are responsible for writing the assembly for the forloop in C: for $(...)$ y$[$i$]=$ h$[$i$]*$ x$[$i$]+$ c;$.$

NOTE: You should put your assembly code within the arraymul$\_$improved$\_$version$1.$c file, as

indicated in asm volatile$($ #include "arraymul$\_$improved$\_$version$1.$c$"$ : $[$h$]"+$r$"$

$($p$\_$h$),...)$; within the improved$\_$version$1()$ function in exercise$2\_1.$c$.$

Your code should use the RISC$-$V V Extension and run with Spike simulator using the specific

configurations $($i$.$e$.,$ vlen$=128,$ elen$=16).$ The vectorized version would improve the execution

efficiency, thanks to the parallel computations done in the vector computation engine.

NOTE: Please do not modify the rest of the header file.

Variables$/$Constants defined in the header files used in this exercise.

Var.$/$Cons$.$ Name Definition

x$[]$ Input array $1$ in arraymul.h

h$[]$ Input array $2$ in arraymul.h

y$[]$ Output array in arraymul.h

improved$\_$version$1\_$cycle$\_$count

Clock cycle in vectorsum$\_$improved$\_$version$1.$c you need

to calculate

cycle$\_$time

The given clock cycle time of the target RISC$-$V processor

running at $2\mathrm{.}6$ GHz

improved$\_$version$1\_$cpu$\_$time

The CPU time in vectorsum$\_$improved$\_$version$1.$c you

need to calculate

arr$\_$size Size of the arrays used in this exercise

student$\_$id

student$\_$id $=$ your$\_$student$\_$id $\%100$

i$.$g$.$ F$12345678$:

student$\_$id $=12345678\%100=78$

Your obtained scores of this exercise is determined by the correctness of your reported performance

data and the efficiency of your developed code against the serial version in the previous exercise.

$1.$ The values of seven counters. $(28\%)$

add$\_$cnt $(4\%)$

sub$\_$cnt $(4\%)$

mul$\_$cnt $(4\%)$

div$\_$cnt $(4\%)$

lw$\_$cnt $(4\%)$

sw$\_$cnt $(4\%)$

others$\_$cnt $(4\%)$

$2.$ The total cycle count $($improved$\_$version$1\_$cycle$\_$count$).(4\%)$

$3.$ The CPU time $($improved$\_$version$1\_$cpu$\_$time$).(4\%)$

$4.$ Achieved speedup. $(14\%)$

If $6<$ speedup, you get $(14$ pt$).$

If $4<$ speedup $<6,$ you get $(9$ pt$)$

If $2<$ speedup $<4,$ you get $(5$ pt$).$

If speedup $<2,$ you get $(0$ pt$).$

The improved$\_$version$1$ function in exercise$2\_1.$c is as follows.

$//$The code snippet for improved$\_$version$1()$ in exercise$2\_1.$c

void improved$\_$version$1()\{$

short int $*$p$\_$h $=$ h;

short int $*$p$\_$x $=$ x;

short int $*$p$\_$y $=$ y;

short int id $=$ student$\_$id;$//$ id $=$ your$\_$student$\_$id $\%100$;

$/*$ original C code

for $($int i $=0$; i $<$ arr$\_$size; i$++)\{$

p$\_$y$[$i$]=$ p$\_$h$[$i$]*$ p$\_$x$[$i$]+$ id;

$\}$

$*/$

asm volatile$($

#include "arraymul$\_$improved$\_$version$1.$c$"//$ Write your code in this file:

arraymul$\_$improved$\_$version$1.$c

: $[$h$]"+$r$"($p$\_$h$),[$x$]"+$r$"($p$\_$x$),[$y$]"+$r$"($p$\_$y$),[$add$\_$cnt$]"+$r$"($add$\_$cnt$),$

$[$sub$\_$cnt$]"+$r$"($sub$\_$cnt$),[$mul$\_$cnt$]"+$r$"($mul$\_$cnt$),[$div$\_$cnt$]"+$r$"($div$\_$cnt$),$

$[$lw$\_$cnt$]"+$r$"($lw$\_$cnt$),[$sw$\_$cnt$]"+$r$"($sw$\_$cnt$),[$others$\_$cnt$]"+$r$"($others$\_$cnt$)$

: $[$id$]"$r$"($id$),[$arr$\_$size$]"$r$"($arr$\_$size$)$

: $"$t$0","$v$0","$v$1","$v$2"$

$)$;

printf$("$

$=====$ Question $2-1=====$

$")$;

printf$("$output: $")$;

for $($int i $=0$; i $<$ arr$\_$size; i$++)\{$

printf$("\%$d $",$ y$[$i$])$;

$\}$

printf$("$

$")$;

printf$("$add counter used: $\%$d

$",$ add$\_$cnt$)$;

printf$("$sub counter used: $\%$d

$",$ sub$\_$cnt$)$;

printf$("$mul counter used: $\%$d

$",$ mul$\_$cnt$)$;

printf$("$div counter used: $\%$d

$",$ div$\_$cnt$)$;

printf$("$lw counter used: $\%$d

$",$ lw$\_$cnt$)$;

printf$("$sw counter used: $\%$d

$",$ sw$\_$cnt$)$;

printf$("$others counter used: $\%$d

$",$ others$\_$cnt$)$;

macro$\_$improved$\_$version$1\_$cycle$\_$count

printf$("$The total cycle count in this program: $\%\mathrm{.}0$f

$",$

improved$\_$version$1\_$cycle$\_$count$)$;

macro$\_$improved$\_$version$1\_$cpu$\_$time

printf$("$CPU time $=\%$f us

$",$ improved$\_$version$1\_$cpu$\_$time$)$;

FILE $*$fp$\_1$;

fp$\_1=$ fopen$("$improved$\_$version$1\_$cpu$\_$time.txt$","$w$")$;

fprintf$($fp$\_1,"\%$f$",$ improved$\_$version$1\_$cpu$\_$time$)$;

fclose$($fp$\_1)$;

float speedup $=0\mathrm{.}0$;

FILE $*$fp$\_2$;

fp$\_2=$ fopen$("$arraymul$\_$baseline$\_$cpu$\_$time.txt$","$r$")$;

fscanf$($fp$\_2,"\%$f$",$ &speedup$)$;

fclose$($fp$\_2)$;

speedup $=$ speedup $/$ improved$\_$version$1\_$cpu$\_$time;

printf$("$The V extension version is $\%$f times faster than the baseline

version

$",$ speedup$)$;

$\}$