c preprocessor functions

**zolfaghar** · 06-16-2016

I am working on converting the shift functionality implemented below into a macro

Code:

#include <stdio.h>
unsigned int shift (unsigned int value, int n)
{
  if ( n > 0 )
      value <<= n;
  else
      value >>= -n;
  return value;
}
int main (void)
{
  unsigned int w1 = 01777777u, w2 = 0444u;
  unsigned int shift (unsigned int value, int n);
  printf ("%o\t%o\n", shift (w1, 5), w1 << 5);
  printf("Here is w1 %o\tHere is w1 left shifted by 1 %o\n", w1, w1 << 1);
  printf ("Right Shift %o\t%o\n", shift (w1, -6), w1 >> 6);
  printf ("\n above line was a right shift of w1:%o by 6", w1);
  printf ("\n%o\t%o\n", shift (w2, 0), w2 >> 0);
  printf ("%o\n", shift (shift (w1, -3), 3));
  return 0;
}

I came up with this....

Code:

#include <stdio.h>
#define  nsign(v,n)  ((n) > (0) ?  (v << n) : (v >> n))
int main (int argc, char **argv)
{
int w1 = 01777777u;
nsign(w1,-5);
printf("%o\n", w1);
}

Which seems to work for positive values of n, but does not seem to work for negative values of n.

How can I improve this? I have looked quite a bit. I read the preprocessor chapter of the book I follow, and a document on C preprocessor for GCC version 2 ( I know it is old, it was the only pdf on the subject I could find ) I also read parts of the GCC documentation. The examples all seem to be similar. I was looking for an example, which had a simple function and its equivalent macro. I also looked into the c library; went through a number of , perhaps half of the .c source files in hope of one macro implementation with if statements and etc. Stackoverflow has a number of very elaborate macros. But they seem to be snippets with the expectation that reader would add the rest I guess. Perhaps after reading all the content, I can understand and implement one them now but they are not exactly what I need. I did try to use what I can from them, but I run into error messages. I wanted to make sure I am not making this overly complex, and ask you for the ideal preprocessor documentation.

So 1- Could you tell me what is the appropriate term I need to search under to find the best documentation on the C preprocess 2- Is what I am doing, the ideal way to go about implementing the shift function, and if not, how can I improve it please? Do I need to implement a compound statement?
Thanks

**Salem** · 06-16-2016

What benefit do you hope to get from this, apart from confusing yourself?

If you say "performance", that's 1980's advice. The idea is redundant nowadays, as most compilers will inline short functions for you.

Your macro doesn't negate n in the negative case.

**zolfaghar** · 06-16-2016

Originally Posted by Salem

What benefit do you hope to get from this, apart from confusing yourself?

If you say "performance", that's 1980's advice. The idea is redundant nowadays, as most compilers will inline short functions for you.

Your macro doesn't negate n in the negative case.

Got it. Ok so just try to make sure I understand some of the stuff, but don't get bogged down with details. Thanks.

**algorism** · 06-16-2016

You need to negate the final n.
You need to give it a better name.
You need to put parens around every use of an argument, and parens around the whole thing (which you did).

And the thing still won't really work properly since it evaluates n more than once (see code below). Real function calls evaluate their inputs only once, and small functions will be inlined when optimized so there's no advantages and definite disadvantages to using a macro for this operation.

Code:

#include <stdio.h>

#define shift(v,n) ((n) >= 0 ? (v) << (n) : (v) >> -(n))

int main(void) {
    unsigned v = 1;
    printf("%u\n", shift(v,        10));
    printf("%u\n", shift(v + v,    10));
    printf("%u\n", shift(v << 10, -10));
    printf("%u\n", shift(v,        3 + 7));

    int n = 9;
    // Here we expect 1024 but we get 2048!
    printf("%u\n", shift(v, ++n));
    // This is because a macro is not really a function
    // but just textual substitution, yielding:
    n = 9;
    printf("%u\n", ((++n) >= 0 ? (v) << (++n) : (v) >> -(++n)));

    return 0;
}

**zolfaghar** · 06-16-2016

Originally Posted by algorism

You need to negate the final n.
You need to give it a better name.
You need to put parens around every use of an argument, and parens around the whole thing (which you did).

And the thing still won't really work properly since it evaluates n more than once (see code below). Real function calls evaluate their inputs only once, and small functions will be inlined when optimized so there's no advantages and definite disadvantages to using a macro for this operation.

Code:

#include <stdio.h>

#define shift(v,n) ((n) >= 0 ? (v) << (n) : (v) >> -(n))

int main(void) {
    unsigned v = 1;
    printf("%u\n", shift(v,        10));
    printf("%u\n", shift(v + v,    10));
    printf("%u\n", shift(v << 10, -10));
    printf("%u\n", shift(v,        3 + 7));

    int n = 9;
    // Here we expect 1024 but we get 2048!
    printf("%u\n", shift(v, ++n));
    // This is because a macro is not really a function
    // but just textual substitution, yielding:
    n = 9;
    printf("%u\n", ((++n) >= 0 ? (v) << (++n) : (v) >> -(++n)));

    return 0;
}

Thanks.

Thread: c preprocessor functions

Thread Tools

Search Thread

Display

c preprocessor functions

Similar Threads

preprocessor

Preprocessor ?

Functions to replace the preprocessor with #defines

preprocessor

Preprocessor Functions

Tags for this Thread