How to display a floating point number by invoking crt_printf?

[MichaelW]

FWIW, this is what the Turbo C++ 3.0 printf does with pi:

Code Select Expand


            1 2345678901234567890
            3.1415926535897932384626...
float       3.14159274101257324000
            3.1415926535897932384626...
double      3.14159265358979312000
            3.1415926535897932384626...
long double 3.14159265358979324000


This is what VC++ 2003 does:

Code Select Expand


            1 2345678901234567890
            3.1415926535897932384626...
float       3.14159274101257320000
            3.1415926535897932384626...
double      3.14159265358979310000
            3.1415926535897932384626...
long double 3.14159265358979310000


And this is what FpuFLtoA does:

Code Select Expand


1 2345678901234567890
3.1415926535897932384626...
3.141592653589793


I think the decision to eliminate long double, despite the existence of well-tested, reliable code to support it, probably indicates that the compiler people did not value those last few digits too highly.

[raymond]

I'll probably try to write algo myself 

That would be an excellent learning experience. Go for it. You could always study the source code in the fpulib wich is only one way to arrive at the result. You may then want to use some other approach but you will at least see which precautions must be taken.

Raymond