Re: [RFC/PATCH] Optimize code based on asserts even in NDEBUG mode

[Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>]

Paolo Bonzini wrote:
> > To followup myself.  I'd rather do this at tree-ssa level, disallowing
> > side-effects in the __builtin_false() arguments (as this would not only
> > compilcate implementation, but also is asking for trouble).
>
>
> My intended approach in tree-ssa would be to expand
>
> __builtin_assert (x)
>
> to
>
> if (__builtin_false () && !x)
>   abort ();
>
> Now the tree optimizers would know nil about __builtin_false () and assume x
> in the rest of the code; at expansion time __builtin_false () would be 0 (like
> when expanding __builtin_constant_p) and the RTL dead code elimination would
> kill the if and more importantly the evaluation of x.

But I suspect having this "special" CFG (how do you call such edges with 
noreturn attribute?) will confuse further high-level tree optimizers, 
such as the loop optimizer?  I.e. is the loop-optimizer happy with

for (int i=i0; i<i1; ++i) {
  if (__builtin_false() && !(i >= mini && i <= maxi))
    abort();
  ...
}

?

Or does it rather like the if() removed before loop optimization?

Richard.