[Bug c++/51559] New: decimal128 operates incorrectly compared to decimal32 and decimal64

[Bug c++/51559] New: decimal128 operates incorrectly compared to decimal32 and decimal64

[mingodad at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51559

             Bug #: 51559
           Summary: decimal128 operates incorrectly compared to decimal32
                    and decimal64
    Classification: Unclassified
           Product: gcc
           Version: 4.6.1
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: mingodad@gmail.com


Created attachment 26092
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26092
Program that demostrates decimal128 fault

Testing the new decimal classes on g++ mingw4.6.1.2 with this program the
decimal32 and decimal64 are showing the expected answer but decimal128 doesn't
all parameters are equal but the result not.

With all 3 decimal classes three instances are created with the following
values:

std::decimal::decimal* dn(.3), dn2(.099), dn3(1000), dn4(201);
dn-=dn2;
dn*=dn3;
cout << "decimal* = "  << (dn==dn4) << " : " << decimal*_to_double(dn) << endl;

Expects "dn==dn4" on all but decimal128 fails.

[Bug c++/51559] decimal128 operates incorrectly compared to decimal32 and decimal64

[mingodad at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51559

--- Comment #1 from Domingo Alvarez <mingodad at gmail dot com> 2011-12-14 23:41:16 UTC ---
Created attachment 26094
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26094
A c program that demonstrate the same problem

Here is a c program that exibits the same problem, so it's not a c++ only
problem probably it's an intrinsic implementation fault of decimal128.

[Bug c++/51559] decimal128 operates incorrectly compared to decimal32 and decimal64

[mingodad at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51559

Domingo Alvarez <mingodad at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID

--- Comment #2 from Domingo Alvarez <mingodad at gmail dot com> 2011-12-15 09:40:35 UTC ---
Doing more tests I discovered that the problem is inherent to using floating
points to initialize decimal128, doing it step by step with integer
initialization gives the expected result.

decimal128 d10(10), d1000(1000), d3(3), d99(99), d201(201);

d3 /= d10); //d3 = 0.3
d99 /= d1000; //d99 = 0.099
d3 -= d99; //d3 = 0.201
d3 *= d1000; // d3 = 201

now d3 == d201

Sorry by the false bug report.

[Bug c++/51559] decimal128 operates incorrectly compared to decimal32 and decimal64

[kubry at terra dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51559

Ganton <kubry at terra dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kubry at terra dot com

--- Comment #3 from Ganton <kubry at terra dot com> ---
> The problem is inherent to using floating points to initialize decimal128

Instead of: 
 - Using a floating point to initialize a decimal128.
 - Dividing integers to initialize a decimal128.
people can use a decimal number to initialize a decimal128. 

In the prior examples, instead of
   std::decimal::decimal128 dn(.3), dn2(.099), dn3(1000), dn4(201);
people can use
   std::decimal::decimal128 dn(.3dl), dn2(.099dl), dn3(1000), dn4(201);