Atmel CAVR-4 Manual De Usuario

CAVR-4

Part 2. Compiler reference

The value of a denormalized number is:

(-1)

S

 * 2

(1-BIAS)

 * 0.Mantissa

where 

BIAS

 is 127 and 1023 for 32-bit and 64-bit floating-point values, respectively.

Note: The IAR CLIB Library does not fully support the special cases of floating-point 
numbers, such as infinity, 

NaN

, and subnormal numbers.

The AVR IAR C/C++ Compiler has two basic types of pointers: function pointers and 
data pointers. 

The size of function pointers is always 16 or 24 bits, and they can address the entire 
memory. The internal representation of a code pointer is the actual address it refers to 
divided by two.

The following function pointers are available:

Data pointers have three sizes: 8, 16, or 24 bits. The following data pointers are 
available:

_ _nearfunc 0–0x1FFFE

2 bytes

signed int

Can be called from any part of the 

code memory, but must reside in 

the first 128 Kbytes of that space.

_ _farfunc

0–0x7FFFFE

3 bytes

signed long No restrictions on code placement.

_ _tiny

1 byte

Data

signed char 0x0–0xFF

_ _near 

2 bytes

Data

signed int

0x0–0xFFFF

_ _far

3 bytes

Data

signed int

0x0–0xFFFFFF

 

(16-bit arithmetics)

_ _huge 

3 bytes

Data

signed long 0x0–0xFFFFFF

_ _tinyflash

1 byte

Code

signed char 0x0–0xFF

_ _flash

2 bytes

Code

signed int

0x0–0xFFFF