https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128069
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target| |x86_64-*-*, aarch64-*-*
Last reconfirmed| |2026-10-21
Ever confirmed|0 |1
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Target Milestone|--- |14.5
Priority|P3 |P2
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. The issue is that we turn
<bb 8> [local count: 1716953]:
# x_1(ab) = PHI <x_8(D)(ab)(2), x_14(ab)(3), x_17(ab)(4), x_19(ab)(5)>
# y_2(ab) = PHI <y_9(D)(ab)(2), y_16(ab)(3), y_18(ab)(4), y_20(ab)(5)>
.ABNORMAL_DISPATCHER (0);
<bb 7> [local count: 1716953]:
<L0>: [non-local]
o[0] = x_1(ab);
o[1] = y_2(ab);
goto <bb 9>; [100.00%]
into
<bb 8> [local count: 1716953]:
# x_1(ab) = PHI <x_8(D)(ab)(2), x_14(ab)(3), x_17(ab)(4), x_19(ab)(5)>
# y_2(ab) = PHI <y_9(D)(ab)(2), y_16(ab)(3), y_18(ab)(4), y_20(ab)(5)>
_26 = {x_1(ab), y_2(ab)};
.ABNORMAL_DISPATCHER (0);
<bb 7> [local count: 1716953]:
<L0>: [non-local]
MEM <vector(2) long int> [(long int *)&o] = _26;
goto <bb 9>; [100.00%]
BB SLP regions should not contain abnormal edges (the edge between 8 and 7 is
one).
I have a patch.
