[Bug rtl-optimization/105314] [12 Regression] ifcvt regression in noce_try_store_flag_mask

[christophm30 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105314

--- Comment #8 from Christoph Müllner <christophm30 at gmail dot com> ---
Yes, I was wrong in my previous comment.
Jakub's patch is of course right.

The transformation in noce_try_store_flag_mask() does:

  x = cond ? 0 else b  // b may be x

==>

  target = cond ? 0 : -1;  // seqz & neg
  x = target & b;          // and

I was misled by the comment above the function,
which only shows the special case when b equals x.

Probably this is a better comment:

/* Convert "if (test) x = 0 else x = y;" to "x = -(test == 0) & y;"  */