Thread: verify MinGW issue

This is more of can you verify my results posting.


I primarily use TDM-GCC ( TDM-GCC ) for my needs but I was interested in this
distro package from: MinGW Distro - nuwen.net


I found different results from the snippit below.
An inquiry suggested I contact the MinGW group.


I then tried using a couple different MinGW versions up to and including the new 4.8.0
They all gave the same results -> 1266


I also tested as:
a c++ example using cl from the Win7.1 sdk
a c++ example using the free Borland 5.5
a BASIC version using PowerBASIC and FreeBasic.
PellesC
tcc


All displayed 1267 except for Borland and MinGW


James






source:

Code:

//***************************************************************
#include <stdio.h>




int     foo (double);
int     main (int,char**);




int foo (double d)
{
  printf("%s% .15G\n","d = ",(double)d);
  return d*100;
}




int main (int argc,char** argv)
{
  int      rv={0};
  rv= foo( 12.67);
  printf("%s% d\n","rv = ",(int)rv);
  return 0;
}
//***************************************************************

This is the result of compiling and running using distro mingw-9.6

Code:

C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>\Mingw\set_distro_paths


C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>gcc --version
gcc (GCC) 4.7.2
Copyright (C) 2012 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.




C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>gcc rv.c -orv.exe


C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>rv
d =  12.67
rv =  1266
*********************************************************************************

this is the result from MinGWTDM

Code:

C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>%MINGW%\bin\gcc --version
gcc (tdm64-1) 4.7.1
Copyright (C) 2012 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.




C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>%MINGW%\bin\gcc rv.c -orv.exe -m32


C:\bc9Adp\RadAsm\Bc9\Examples\bc9CppExamples>rv
d =  12.67
rv =  1267

@grumpy -> excellent explanation thank you.

Now if I change d to a float instead of a double, 1267 is displayed across the board with all compilers I tried.

James

Thanks everyone.


This is my latest test piece. When compiled with the newest 32bit gcc 4.8.0 only the first one displays 1266


James

Code:

#include <stdio.h>
#include <math.h>
int     foo (double);
int     main (int,char**);


int foo (double d)
{
  printf("%s% .15G\n","d = ",(double)d);
return (d*100);      // 1266
//return ceil(d*100);  // 1267
//return floor(d*100); // 1267
//return trunc(d*100); // 1267
//return round(d*100);   // 1267
}


int main (int argc,char** argv)
{
  int      rv={0};
  double d =12.67;
  rv= foo( d);
  printf("%s% d\n","rv = ",(int)rv);
  return 0;
}

Originally Posted by jcfuller

Now if I change d to a float instead of a double, 1267 is displayed across the board with all compilers I tried.

I tried your last test -- on GCC 4.7.2 (x86_64-linux-gnu, Ubuntu) it gives me 1267 for doubles and floats. If I change to "long double" it gives me 1267 for 12.67L but 1266 for 12.67 (i.e. assigning a double constant to a long double variable).

Although floating point types can and will drop accuracy on the floor, this example makes it look worse than it is. If you ask printf for 50 digits instead of 15 it looks more forgivable.

Code:

d =  12.66999999999999992894572642398998141288757324218750
rv =  1266

d =  12.66999999999999992894572642398998141288757324218750
rv =  1267

Thank you all for your responses.
I do have a better understanding of the what is happening.
I also took the question to another venue I frequent :
fpu example

I ran into this scenario while working on example code for bc9: bc9 home and got a bit carried away.

Most users probably have even less c/c++ experience that I do (which is none too great) so I was looking for a way (if possible) to return the same value for all compilers supported by bc9.

James

I believe I found the answer thanks to MichaelW on the masm32 forum from the link in the last msg.
I am surprised no one here was aware of this???

From MinGw float.h


MSVCRT.dll _fpreset initializes the control register to 0x27f,
the status register to zero and the tag word to 0FFFFh.
This differs from asm instruction finit/fninit which set control
word to 0x37f (64 bit mantissa precison rather than 53 bit).
By default, the mingw version of _fpreset sets fp control as
per fninit. To use the MSVCRT.dll _fpreset, include CRT_fp8.o when
building your application.


James