Everyone,
> I'm using msp430-gcc 3.2.2, which upto now worked fine, but
> is giving me strange behaviour now. The following code (which when compiled
> identically for the PC works perfectly), produces incorrect results when
> compiled by msp430-gcc -O2:
>
> #define GETBIT(d,n) (((d)>>(n))&1)
> // calculates 4 parity bits for (8,12) hamming
> unsigned char hamming_parity(unsigned char d)
> {
> unsigned char HP1,HP2,HP3,HP4;
>
> // calculate parity bits
> HP1 = GETBIT(d,0) ^ GETBIT(d,1) ^ GETBIT(d,3) ^ GETBIT(d,4) ^ GETBIT(d,6);
> HP2 = GETBIT(d,0) ^ GETBIT(d,2) ^ GETBIT(d,3) ^ GETBIT(d,5) ^ GETBIT(d,6);
> HP3 = GETBIT(d,1) ^ GETBIT(d,2) ^ GETBIT(d,3) ^ GETBIT(d,7);
> HP4 = GETBIT(d,4) ^ GETBIT(d,5) ^ GETBIT(d,6) ^ GETBIT(d,7);
>
> // combine into parity nibble
> return (HP4<<3)|(HP3<<2)|(HP2<<1)|HP1;
> }
For those that experience the same bug, I thought I'd post the way I was able
to get round this: simply encapsulate the macro in a function, like this:
#define BIT(d,n) getbit(d,n)
static unsigned char inline getbit(unsigned char d, unsigned char n) {return
(((d)>>(n))&1);}
// calculates 4 parity bits for (8,12) hamming
unsigned char hamming_parity(unsigned char d)
{
unsigned char P1,P2,P3,P4;
// calculate parity bits
P1 = BIT(d,0) ^ BIT(d,1) ^ BIT(d,3) ^ BIT(d,4) ^ BIT(d,6);
P2 = BIT(d,0) ^ BIT(d,2) ^ BIT(d,3) ^ BIT(d,5) ^ BIT(d,6);
P3 = BIT(d,1) ^ BIT(d,2) ^ BIT(d,3) ^ BIT(d,7);
P4 = BIT(d,4) ^ BIT(d,5) ^ BIT(d,6) ^ BIT(d,7);
// combine into parity nibble
return (P4<<3)|(P3<<2)|(P2<<1)|P1;
}
Maybe someone with the needed time and expertise can have a look at what causes
the compiler to emit the bad code (see original post) in the first place? Or is
this bug maybe already fixed in a more recent version of the compiler?
greetings,
Tom
