On 09/06/2011 08:50 PM, malc wrote: > Correct me if i'm wrong, previously the code worked like this: > > mov tmp, 0 > sub tmp, 1 > sar r, r, tmp > > Still UB as far as TCG is concerned but since no immediates are involved > things worked, now, with constant folding, we are asked to sar by -1 > directly.
You are exactly correct. That's why I thought my patch to re-arrange the order of operations and only perform the subtraction inside the %cl != 0 test was a good idea. No point in performing the shift if we're not going to use the result. > I did that when first hit this problem, but decided not to push it. I think pushing it is a good idea. Just because the result is not defined, according to tcg/README, is no reason to SIGILL. r~
