On Jun 21, 2012, at 6:00 PM, Christopher Li wrote:
>
> I like this series a lot. Do you have an update version of the patch?
I will resend a revised version of these patches.
With ->ctype in pseudo, we might have to disable several simplifications
in simplify.c and cse.c for type consistency. One such example is:
unsigned int foo(unsigned long long x)
{
unsigned int tmp = x & 0x1fffffff;
return tmp;
}
sparse emits:
and.64 %r2 <- %arg1, $0x1fffffff
ret.32 %r2
%r2 seems to be interpreted as both 64 and 32 bits, which will also
trouble backends.
To fix the inconsistency, we may need to disable some simplification:
https://github.com/xiw/sparse/commit/7072002
Then we have:
and.64 %r2 <- %arg1, $0x1fffffff
cast.32 %r3 <- (64) %r2
ret.32 %r3
Does this make sense? Or should we just provide an option to turn
off all sparse simplifications, since backends usually have their own
optimization passes?
- xi
--
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
