Go Premium for a chance to win a PS4. Enter to Win

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 1964
  • Last Modified:

pow(10,x) implementation!

hi,

  could anybody tell me, how can i implement pow(10,x) function, where x can be double? ofcourse, i want to minimize use of library functions to speedup the process?
0
Vikram_B
Asked:
Vikram_B
  • 2
  • 2
  • 2
1 Solution
 
sunnycoderCommented:
here is the glibc implementation of pow() (which I would not strain to comprehend)

# define __TGMATH_BINARY_REAL_IMAG(Val1, Val2, Fct, Cfct) \
     (__extension__ ({ __tgmath_real_type ((Val1) + (Val2)) __tgmres;            \
                   if ((sizeof (__real__ (Val1)) > sizeof (double)            \
                      || sizeof (__real__ (Val2)) > sizeof (double))    \
                     && __builtin_classify_type (__real__ (Val1)            \
                                           + __real__ (Val2))     \
                        == 8)                                    \
                   {                                          \
                     if (sizeof (__real__ (Val1)) == sizeof (Val1)      \
                         && sizeof (__real__ (Val2)) == sizeof (Val2))  \
                       __tgmres = __tgml(Fct) (Val1, Val2);            \
                     else                                          \
                       __tgmres = __tgml(Cfct) (Val1, Val2);            \
                   }                                          \
                   else if (sizeof (__real__ (Val1)) == sizeof (double)   \
                        || sizeof (__real__ (Val2)) == sizeof(double) \
                        || (__builtin_classify_type (__real__ (Val1)) \
                            != 8)                              \
                        || (__builtin_classify_type (__real__ (Val2)) \
                            != 8))                              \
                   {                                          \
                     if (sizeof (__real__ (Val1)) == sizeof (Val1)      \
                         && sizeof (__real__ (Val2)) == sizeof (Val2))  \
                       __tgmres = Fct (Val1, Val2);                  \
                     else                                          \
                       __tgmres = Cfct (Val1, Val2);                  \
                   }                                          \
                   else                                          \
                   {                                          \
                     if (sizeof (__real__ (Val1)) == sizeof (Val1)      \
                         && sizeof (__real__ (Val2)) == sizeof (Val2))  \
                       __tgmres = Fct##f (Val1, Val2);                  \
                     else                                          \
                       __tgmres = Cfct##f (Val1, Val2);                  \
                   }                                          \
                   __tgmres; }))


if you feel lost then there is always the simpler one

result = 1;
for ( i = 0;  i < x; i++ )
     result = result  * 10;

;o)
0
 
sunnycoderCommented:
a little study shows that pow is implemented as a macro and not as a function... if you use pow(), your execution time would not be adversely affected... moreover, library implementation of pow() is bound to be more efficient what I or you would write... So I would say --- use pow()
0
 
Vikram_BAuthor Commented:
life is not always straignt-forward. i'm implementing the pow function on a dsp chip (TI'x 64xx) which doesn't have floating pt unit. thus, using libraty pow function is consuming lotta cpu cycles.
0
Technology Partners: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

 
SethHoytCommented:
You might consider changing bases to either 2 or e by precomputing the logarithm of 10 for the desired base. Many systems have an optimized implementation for exponentiation base e, and base 2 can be done quickly for powers in binary form using the square and multiply method.

To transform the base to 2, use the identity:

10^x = (2^lg(10))^x = 2^(lg(10)*x)

where lg() is the logarithm base 2. Typically, you will need to compute log(10)/log(2) using a different base, but this is done only once and stored for later use, so efficiency is not important here. The transformation is similar for base e, but unless a built-in function exists for exponentiation base e, you should use base 2 for this.

So the idea is to compute r = log(10)/log(2) one time, either globally or offline, then find 2^(r*x).


-Seth
0
 
SethHoytCommented:
You can find the square and multiply method here:

http://www.math.ucalgary.ca/~kjell/papers/hardware/gordon98survey.pdf

-Seth
0
 
Vikram_BAuthor Commented:
thanx SethHoyt,

 that should work :)
0

Featured Post

Industry Leaders: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

  • 2
  • 2
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now