Atmel CAVR-4 Manual De Usuario

CAVR-4

Basic settings for project configuration

Reference Guide

Your program may use only one memory model at a time, and the same model must be 
used by all user modules and all library modules. 

The following table summarizes the characteristics for each memory model:

For more details about memory models, see Memory models, page 17.

Floating-point values are represented by 32- and 64-bit numbers in standard IEEE754 
format. By enabling the compiler option 

--64bit_doubles

, you can choose whether 

data declared as 

double

 should be represented with 32 bits or 64 bits. The data type 

float

 is always represented using 32 bits.

The AVR IAR C/C++ Compiler is a state-of-the-art compiler with an optimizer that 
performs, among other things, dead-code elimination, constant propagation, inlining, 
common sub-expression elimination, and precision reduction. It also performs loop 
optimizations, such as induction variable elimination.

You can decide between several optimization levels and two optimization goals—size 
and speed. Most optimizations will make the application both smaller and faster. 
However, when this is not the case, the compiler uses the selected optimization goal to 
decide how to perform the optimization.

The optimization level and goal can be specified for the entire application, for individual 
files, and for individual functions. In addition, some individual optimizations, such as 
function inlining, can be disabled.

For details about compiler optimizations, see Controlling compiler optimizations, page 
122. For more information about efficient coding techniques, see the chapter Efficient 
coding for embedded applications.

Tiny

-v0

, -v1, -v2,

-v3

, -v5

_ _tiny

_ _tiny

≤ 256 bytes

Small

-v1

, -v3, -v4,

-v5

, -v6

_ _near

_ _near

≤ 64 Kbytes

Large

-v4

, -v6

_ _far

_ _far

≤ 16 Mbytes