[ILUG] gcc optimisation weirdness?

[Paul Jakma]

On Wed, 10 May 2000, Kenn Humborg wrote:

> 
> Very odd...  I'd expect that the statement
> 
>    (f[i])++
> 
> would not be legal.  Is (var) a legal lvalue?  Aha... info gcc

well i initially didn't have brackets. I added them because it seemed
that with:

frequency[input+1]++;

gcc was somehow mis-optimising the ++, so i added the () to try to make it
explicit that i wanted frequency[input-1] evaluated first, and the
increment to be applied to frequency[..].

[snip gcc info]

i'm not doing anything funky like these examples above. i'm just trying to
be explicit about the ordering wrt to the ++ operator.

> > PS: is there anyway to initialise an array at declaration time, rather
> > than iterate thru the array and set each element manually? eg something
> > like: int array[9]=0;
> 
> You mean like:
> 
>    int array[9]={0, 1, 2, 3, 4, 3, 2, 1, 0};
> 
> That is standard C.  

doh! i'm rusty... just coulnd't remember it.

> GCC also has some extensions that allow
> you to do fancied stuff with initializers.

the : operator, right? don't remember exactly how it works, but something
like:

struct foo {
	int var : 0;
	char bar : NULL;
} foobar;

it used in a few places in the linux source.

> Take a look at
> the C and C++ Extensions sections in the GCC info pages.
> 

still can't understand exactly where "gcc test.c" is fscking up or why,
when "gcc -Ox" works as expected.

How's your x86 asm? :) 

> Later,
> Kenn
>