Hello,
I face a strange problem with gcc that I never faced with other compilers.
If for example, I have the following definitions :
In an include file :
# define BASE 0xF000# define REGISTER BASE + 0x02# define UBYTE unsigned char
In a code file :
*((UBYTE *)(REGISTER) = 0x00;
...
*((UBYTE *)(REGISTER) = 0xFF;
When i look at the assembler code, the address is calculated each time (0xF000 + 0x02) instead of being calculated at the compilation time.
Although both are literals ...
This involves additions, register transfers and is 2-3 times longer than what I can do by hand ..
This is very strange as both values are known at compilation time. Even if we use optimisation (-O2 or O3), it does nothing toward an optimisation which is not very difficult to carry out.
Did you faced such issues ?
Any idea or any optimisation flag to be used that you know about ?
With best regards,
Thierry
