300pt 80x86 ASM question

Hand crafted asm can be over 100 times faster. Even bad asm will get 10% faster, whats the point?
I'll give it a spin, but I know the result already...

Here is the best I can do in short order!

#include<memory> //lazy me!        
#include<cstdlib> //rand()
#include<ctime>
using namespace std;

double jtimer;
inline void start_time()
{
     jtimer = (double)(clock());
}

inline void elapsed_time(char *ch)
{
     jtimer = clock() - jtimer;
     jtimer = jtimer/CLOCKS_PER_SEC;
     printf("%s time %f seconds\n",ch, jtimer);    
}


static unsigned char ascii[255]; //map '.' into the required ascii fields
static unsigned char *data;    
//the data. you did not say well written, you said fast!

inline void f()
{
    printf("%x %x %x %x %x %x %x %x-%x %x %x %x %x %x %x %x",
        data[0],data[1],data[2],data[3],data[4],data[5],data[6],
        data[7],data[8],data[9],data[10],data[11],data[12],data[13],
        data[14],data[15]);
    printf("  %c%c%c%c%c%c%c%c%c%c%c%c%c%c%c%c\n",
        ascii[data[0]],
        ascii[data[1]],
        ascii[data[2]],
        ascii[data[3]],
        ascii[data[4]],
        ascii[data[5]],
        ascii[data[6]],
        ascii[data[7]],
        ascii[data[8]],
        ascii[data[9]],
        ascii[data[10]],
        ascii[data[11]],
        ascii[data[12]],
        ascii[data[13]],
        ascii[data[14]],
        ascii[data[15]]);
}
int d[4];
int main()
{
    memset(ascii,'.',255);
    short i;
    for(i = 32; i < 128; i++)
        ascii[i] = i;
   
       
    data = (unsigned char *)d;
    i = 1;
    start_time();
    while(i++)
    {
        d[0] = rand();
        d[1] = rand();
        d[2] = rand();
        d[3] = rand();
        f();
    }
    elapsed_time("");
    /*
     I ran re-directed to a file
     b > a.txt
     the result was less than
     0.98 seconds for 65535 of them
     on a 1000 mhz AMD processor, win 2k.
     all optimizations tweaked for this app
     including:
     no c++ exceptions
     intel 386 exe format for speed
     inline expansion
     no debug stuff
     maximize speed chosen
     this can be faster still but for
     30 min is all I can afford for today.
     */
    return(3);
}

hi jonnin!
Thanks for your submission.  I posted the core of it into the ASM question and will post the timimg results tomorrow.  Check it out, and please post in that thread rather than this one.  Thanks!

-- Dan

Hi Pavlik,
I posted your code and timing of it on the other thread.  You are looking good.

https://www.experts-exchange.com/jsp/qManageQuestion.jsp?ta=assembly&qid=20272901

You may want to go there and subscribe to that Q so you can get a notif of the final outcome.

-- Dan

The original Assembly TA thread has been continued here:
   http:Q.20280946.html
and then here:
   http:Q.20283475.html
Join in the discussion, we are having fun!

-- Dan

Hi Pavlik,
I added 100 points to this Q.  Pleaase accept them with my complements for your excellent C++ code.  Yours beet everyone until we started messing around with MMX operations.

You might want to check in on the thread.  'able' built a version with a 143MB lookup table.  It's all too, too, tooo... well what the heck.  

-- Dan