Atmel CAVR-4 Manual De Usuario

CAVR-4

Part 2. Compiler reference

In an initializer, a pointer constant value may be cast to an integral type if the integral 
type is large enough to contain it.

In the following example, it is assumed that pointers to 

_ _near

 and 

_ _huge

 are 16 and 

32 bits, respectively. The first initialization is correct, because it is possible to cast the 
16-bit address to a 16-bit 

unsigned

 

short

 variable. However, it is illegal to use the 

32-bit address of 

b

 as an initializer for a 16-bit value.

_ _near int a;

_ _huge int b;

unsigned short ap = (unsigned short)&a;  /* Correct */

unsigned short bp = (unsigned short)&b;  /* Error */

Floating-point constants can be given in hexadecimal style. The syntax is 

0xMANTp{+|-}EXP

, where 

 is the mantissa in hexadecimal digits, including an 

optional 

.

 (decimal point), and 

 is the exponent with decimal digits, representing an 

exponent of 2.

0x1p0

 is 

1

0xA.8p2

 is 

10.5*2^2

To use the 

bool

 type in C source code, you must include the file 

stdbool.h

. (The 

bool

 

data type is supported by default in C++.)

In ISO/ANSI C, it is illegal to take the address of a variable specified as a register 
variable.

The AVR IAR C/C++ Compiler allows this, but a warning is issued.

Should the size or sign specifiers be duplicated (for example, 

short

 

short

 or 

unsigned

 

unsigned

), an error is issued. 

long

 

float

 is accepted as a synonym for 

double

.