[Boost-bugs] [Boost C++ Libraries] #12408: cpp_dec_float_50 x = x*x*x gives wrong result!!!

Subject: [Boost-bugs] [Boost C++ Libraries] #12408: cpp_dec_float_50 x = x*x*x gives wrong result!!!
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2016-08-23 11:42:26

#12408: cpp_dec_float_50 x = x*x*x gives wrong result!!!
---------------------------------------+------------------------------
 Reporter:  Dave Nalepa <denalepa1@…>  |       Type:  Bugs
   Status:  new                        |  Milestone:  To Be Determined
Component:  None                       |    Version:  Boost 1.61.0
 Severity:  Problem                    |   Keywords:
---------------------------------------+------------------------------
 I'm a beginner with C++ and with boost so I won't be surprised if I'm
 making some blunder but in case there's a real problem here, here goes:
 Using the multiprecision cpp_dec_float_50 I get a enormous error when I
 cube a variable.
 I'm using Microsoft VS2015 Community and making a console program (since
 my attempts at using boost with a cli program failed miserably).
 The core code (I think) is
 #include <boost/multiprecision/cpp_dec_float.hpp>
 using boost::multiprecision::cpp_dec_float_50;
 // both at the global level = before main(){...
 I set f2 to this and then use the line
 f2 = f2*f2*f2; //to compute (f2)³
 when f2 = 1.0335055959... I get the new f2 as 1.140909... which isn't even
 close to the correct value of 1.10392227...
 When I use a temporary variable, f3, it works just fine
 f3= f2*f2*f2 // f3 = 1.1039...
 It is probably some error the compiler is making, I don't know. This is
 obviously a worrisome bug. Simple arithmetic shouldn't give results that
 are off by 3% !!
 I'm not skilled enough in VS2015 to know what else to tell you about the
 switches that are or are not selected; I compile and run in the VS IDE.
 //===================================== here's the entire program (stub)
 // I've shortened 1.03559... but I still get the wrong value.
 #include <iomanip>
 #include <iostream>
 #include <boost/multiprecision/cpp_dec_float.hpp>

 using boost::multiprecision::cpp_dec_float_50;

 int main()
 {
         cpp_dec_float_50 f2 = (cpp_dec_float_50)(1.03559);
         f2 = f2*f2*f2;
         std::cout << "f2 = " << f2 << std::endl;
 // gives me 1.15.... which is wrong should be 1.1106...
 // it is actually the value of f2*f2*f2*f2, or close to it.
     return 0;
 }

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/12408>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.