Compiler/CC2640: number of decimal support in float variable

Dilbert K

Expert 1210 points

Part Number: CC2640

Tool/software: TI C/C++ Compiler

I would like to confirm how many decimal point can be supported for this TI platform ?

I am using this for GPS location and try to use 6 decimal.

For the positive number, it is fine. But for negative, it show incorrect decimal value.

e.g.

float value;

value = -70.567890;

In CCS debugger, when I examine the value, it display as

Default:-70.5678864
Hex:0xC28D22C2
Decimal:-70.5678864
Octal:030243221302
Binary:11000010100011010010001011000010b

over 8 years ago

0 Archaeologist over 8 years ago

TI__Guru* 84285 points

CC2640 has an ARM Cortex-M3. For all variants of ARM, "float" is IEEE-32 and "double" is IEEE-64. The hardware doesn't really keep track of floating-point numbers as decimal, but the header file float.h will tell you the approximate decimal answer. FLT_DIG is 6, so you can expect about 6 significant decimal digits for a "float." DBL_DIG is 15, so you can expect about 15 significant decimal digits for a "double."

The value -70.567890 cannot be represented exactly in 32-bit float, so the compiler gives you the closest value, -0x1.1a4584p+6, or 0xc28d22c2, or (approximately)-70.567886

0 user347219 over 8 years ago

Expert 1240 points

I find these converts helpful sometimes:
www.h-schmidt.net/.../IEEE754.html
www.exploringbinary.com/.../

And at times I will find myself referring to:
docs.oracle.com/.../ncg_goldberg.html

0 Dilbert K over 8 years ago in reply to Archaeologist

Expert 1210 points

Thanks for great info and confirmation.... in order to make sure all 6 decimal are 100%, I must use the "double" to be sure.

0 Archaeologist over 8 years ago in reply to Dilbert K

TI__Guru* 84285 points

You're never guaranteed 100% agreement; some decimal values, such as 1/3, are not exactly expressible in either float or double.

Don't forget, you must also count the digits before the decimal point. Consider 70.567890. Six decimal digits would be 70.5679

Yes, it seems like you need double precision.

Code Composer Studio™︎

Code Composer Studio forum

Compiler/CC2640: number of decimal support in float variable