https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91275
Segher Boessenkool <segher at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |wschmidt at gcc dot gnu.org
--- Comment #4 from Segher Boessenkool <segher at gcc dot gnu.org> ---
It is caused by the "swaps" (p8swap) pass. Before this pass we have:
5: load r122
6: r123 = swap r122
7: r121 = swap r123
8: load r125
9: r126 = swap r125
10: r124 = swap r126
11: r117 = vpmsumd r121, r124
12: r127 = vec_select r117, 1 # this is the high dword, 0 in hardware
13: r128 = vec_select r117, 0 # this is the low dword, 1 in hardware
14: load r129
15: r5 = r127
16: r4 = r128
17: r3 = r129
18: call printf
The swaps pass replaces 7 and 10 by plain moves.
