https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120629
--- Comment #18 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
With --param=ranger-debug=all I see
21 range_of_expr(_57) at stmt __builtin_memset (_60, 0, _57);
22 range_of_stmt (_57) at stmt _57 = _56 * 4;
23 ROS dependence fill
ROS dep fill (_57) at stmt _57 = _56 * 4;
ROS dep fill (_56) at stmt _56 = _55 + 1;
ROS dep fill (_55) at stmt _55 = (sizetype) _54;
24 range_of_expr(_54) at stmt _55 = (sizetype) _54;
TRUE : (24) range_of_expr (_54) [irange] unsigned int [7, 7]
Registering value_relation (_55 pe32 _54) (bb11) at _55 = (sizetype) _54;
GLOBAL : UPDATE cache for _55 in BB 11 : successors : : No updates!
25 range_of_expr(_55) at stmt _56 = _55 + 1;
TRUE : (25) range_of_expr (_55) [irange] sizetype [7, 7] MASK
0x7 VALUE 0x0
Registering value_relation (_56 > _55) (bb11) at _56 = _55 + 1;
GLOBAL : UPDATE cache for _56 in BB 11 : successors : : No updates!
FALSE : (23) ROS (_57)
26 range_of_expr(_56) at stmt _57 = _56 * 4;
TRUE : (26) range_of_expr (_56) [irange] sizetype [8, 8] MASK 0xf
VALUE 0x0
GLOBAL : UPDATE cache for _57 in BB 11 : successors : : No updates!
TRUE : (22) range_of_stmt (_57) [irange] sizetype [32, 32] MASK 0x3c
VALUE 0x0
_54 seems to have global range [0, 7] though.
