Re: Should we use boost::math::log1p/expm1 instead of rtl_math_log1p/expm1 ?

On 09/15/2014 03:38 PM, julien2412 wrote:
> Hello,
>
> Following
> http://cgit.freedesktop.org/libreoffice/core/commit/?id=a3ad01dc1e07da21182077bd899094fd73f57714,
> should  it worth it to replace rtl_math_log1p/rtl_math_expm1 and related by
> boost::math::log1p/expm1 ?
> Or should we let the code as it is right now?
>
> Since we already use Boost, I thought it could be interesting to use it even
> more, waiting by a even more C++11 compatibility (since this standard
> includes these 2 functions).
>
> Julien
>
>
>
> --
> View this message in context: 
> http://nabble.documentfoundation.org/Should-we-use-boost-math-log1p-expm1-instead-of-rtl-math-log1p-expm1-tp4122334.html
> Sent from the Dev mailing list archive at Nabble.com.
> _______________________________________________
> LibreOffice mailing list
> LibreOffice@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libreoffice

Julien, I am curious if you have done any testing. On a whim, I ran the 
following on my 64-bit Fedora machine using g++.

#include <boost/math/special_functions/log1p.hpp>
#include <boost/math/special_functions/expm1.hpp>

#include <iostream>
#include <math.h>

int main()
{
  double min_x = -0.2;
  double max_x = 0.2;
  long max_its = 10000000;
  double delta_x = (max_x - min_x) / (max_its - 1);
  double x;
  double d;
  long i;
  for (i=0; i<max_its; ++i) {
    x = min_x + i * delta_x;
    d = fabs(log1p(x) - boost::math::log1p(x));
    if (d > 0.0) {
      std::cout << "1. diff(" << x << ") = " << d << std::endl;
    }
  }

  for (i=0; i<max_its; ++i) {
    x = min_x + i * delta_x;
    d = fabs(expm1(x) - boost::math::expm1(x));
    if (d > 0.0) {
      std::cout << "2. diff(" << x << ") = " << d << std::endl;
    }
  }
  std::cout << "Finished!" << std::endl;
  return 0;
}

No differences were found in this little test. Then again, on Fedora, 
Boost may be configured to use the built-in implementation, I don't 
really know. I did not check our source code, but, it might reduce the 
LO code complexity a little bit, that might be a gain.

--
Andrew Pitonyak
My Macro Document: http://www.pitonyak.org/AndrewMacro.odt
Info:  http://www.pitonyak.org/oo.php