https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127100
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
IFN_LEN_LOAD callls in *.optimized dump in r16-5983 are
vect__14.19_56 = .LEN_LOAD (&__var_1_matmul, 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0 }, 1, -1);
vect__19.22_102 = .LEN_LOAD (&a, 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0 }, 1, -1);
vect__14.19_31 = .LEN_LOAD (&__var_1_matmul, 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0 }, 1, -1);
vect__19.22_28 = .LEN_LOAD (&MEM <integer(kind=1)[4]> [(void *)&a + 2B], 8B,
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, 1, -1);
vect__14.19_7 = .LEN_LOAD (&MEM <integer(kind=1)[4]> [(void *)&__var_1_matmul
+ 2B], 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, 1, -1);
vect__19.22_25 = .LEN_LOAD (&a, 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0 }, 1, -1);
vect__14.19_61 = .LEN_LOAD (&MEM <integer(kind=1)[4]> [(void
*)&__var_1_matmul + 2B], 8B, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
}, 1, -1);
vect__19.22_40 = .LEN_LOAD (&MEM <integer(kind=1)[4]> [(void *)&a + 2B], 8B,
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, 1, -1);
so guess those are what the r16-5984 folds to something else.
