https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98478
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
If the selftests are the only problem, I don't see a reason why the self-test
shouldn't be skipped for V*BImode or perhaps skipped for V*BImode with some
extra conditions.
E.g. skip the test if the scalar mode doesn't have mov_optab insn?
static void
test_vector_ops ()
{
for (unsigned int i = 0; i < NUM_MACHINE_MODES; ++i)
{
machine_mode mode = (machine_mode) i;
- if (VECTOR_MODE_P (mode))
+ if (VECTOR_MODE_P (mode)
+ && optab_handler (mov_optab, GET_MODE_INNER (mode)) !=
CODE_FOR_nothing)
?
I mean, simplify-rtx.c will not really try to do the VEC_SELECT etc.
simplifications on V*BImode if
something like that isn't present in the IL before the simplification, and I
seriously doubt it can
appear there when one even can't have a BImode pseudo and move it around.
