Bit Reversal

Not good lenthy
Comapred toi macro ....!!
One can ewrite a macro to doall above in single shot...!!

Not good lenthy
Comapred toi macro ....!!
One can ewrite a macro to doall above in single shot...!!

You obviously don't know what you are talking about.

I know this is old, but anyway...

I've always preferred a simple variation of bar():

unsigned char rev( unsigned char b ) {
  unsigned char result = 0;
  while (b) {
    result <<= 1;
    result  |= b % 2;
    b      >>= 1;
    }
  return result
  }

This method has the nice property that it is 100% portable.

Nice and informative article
and IMO for some thing like bit order reversal we can not think much of portability.
One suggestion, we can combine two algos here:
1. Calculate the bit reversed values using the zot()
2. Put these values in array indexes as in fum()

Now one can access the reversed values using the array indexes.

~Trib

thanks..
but i did not understand anything (i mean purpose)inside

for ( i = 0; i < sizeof(function)/sizeof(*function); ++i )
{

loop other than

printf("%s(%X) = %X", name [ i ], 0x5C, function [ i ] (0x5C));

can you please tell me why you have measured the performance and how it is done?

but i did not understand anything (i mean purpose)inside

I don't understand what you don't understand (perhaps consider mentioning what it is you don't understand?).

#define CYCLES 100000000

int main(void)
{
   const char *name[] = { "foo", "bar", "baz", "qux", "zot", "fum" };
   unsigned char (*const function[])(unsigned char) =
   {
      foo, bar, baz, qux, zot, fum
   };
   size_t i,j;
   fflush(stdout);
   for ( i = 0; i < sizeof(function)/sizeof(*function); ++i )
   {
      clock_t end, start;
      printf("%s(%X) = %X", name [ i ], 0x5C, function [ i ] (0x5C));
      fflush(stdout);
      start = clock();
      for ( j = 0; j < CYCLES; ++j )
      {
         function [ i ] (j);
      }
      end = clock();
      printf(", elapsed time = %g\n", (end - start) / (double)CLOCKS_PER_SEC);
   }

   return 0;
}

Here we declare two variable to receive the result of calling the clock() function:

clock_t end, start;

The variable start will be before we loop through each function being tested (the start). The variable end will be after we loop through each function being tested (the end).

The flush is used to ensure that we see the preceding printf. Otherwise, it might not appear as we are looping, which I find somewhat annoying.

fflush(stdout);

http://c-faq.com/stdio/fflush.html

As mentioned, before we start the loop, we get the current clock() value.

start = clock();

Then we have a loop that executes the function the defined number of cycles.

for ( j = 0; j < CYCLES; ++j )
{
    function [ i ] (j);
}

The variable function is perhaps the most complicated part if you are not yet familiar with function pointers; function is just an array of pointers to the functions defined further up in the code:

unsigned char (*const function[])(unsigned char) =
{
   foo, bar, baz, qux, zot, fum
};

That is, when the main loop variable i is 0, the function being called is foo . When i is 1, the function being called is bar . And so on.

Each of these functions is being passed a parameter, which is the current value of j . I wrote this code so long ago, I'm not sure whether I intentionally meant to use the implicit truncation to an unsigned char, or whether I simply didn't care about the passed parameter any further than each call to the function was passed something and that it didn't really matter for the timing.

After the loop is done, we get the ending clock() value.

end = clock();

And then we print the results.

printf(", elapsed time = %g\n", (end - start) / (double)CLOCKS_PER_SEC);

The cast to double is done because some machines might end have integral clock_t and CLOCKS_PER_SEC values. I wanted the number of seconds in a floating-point value for elapsed time.

Hi ,

Thank you for the detailed explanation +1

Each of these functions is being passed a parameter, which is the current value of j . I wrote this code so long ago, I'm not sure whether I intentionally meant to use the implicit truncation to an unsigned char, or whether I simply didn't care about the passed parameter any further than each call to the function was passed something and that it didn't really matter for the timing.

This is was exactly where i got confused(passing a big value to unsigned char) more than the case of function pointers. I understood the use of function pointers in the printf() statement .

the case of fflush is interesting. one more question ..
http://c-faq.com/ - is this site highly dependable?
I have seen reference to this site at more than one place from this forum. this is just a doubt since i am looking for dependable resources on web to study C.

http://c-faq.com/ - is this site highly dependable?

Yes. It is the FAQ from the usenet newsgroup comp.lang.c, which is probably the oldest "forum", and it is populated by a group of the most authoritative folks out there. The only issue with this FAQ is that it has aged a bit.

i am looking for dependable resources on web to study C.

Steve Summit, the FAQ maintainer, also has this:
http://c-faq.com/~scs/cclass/notes/top.html