On Mon, 8 Feb 2021 at 03:28, Richard Henderson <richard.hender...@linaro.org> wrote: > > The three vector shift by vector operations are all implemented via > expansion. Therefore do not actually set TCG_TARGET_HAS_shv_vec, > as none of shlv_vec, shrv_vec, sarv_vec may actually appear in the > instruction stream, and therefore also do not appear in tcg_target_op_def. > > Signed-off-by: Richard Henderson <richard.hender...@linaro.org> > --- > tcg/arm/tcg-target.opc.h | 3 ++ > tcg/arm/tcg-target.c.inc | 61 +++++++++++++++++++++++++++++++++++++++- > 2 files changed, 63 insertions(+), 1 deletion(-)
> + switch (opc) { > + case INDEX_op_shlv_vec: > + /* > + * Merely propagate shlv_vec to arm_ushl_vec. > + * In this way we don't set TCG_TARGET_HAS_shv_vec > + * because everything is done via expansion. > + */ > + v2 = temp_tcgv_vec(arg_temp(a2)); > + vec_gen_3(INDEX_op_arm_ushl_vec, type, vece, tcgv_vec_arg(v0), > + tcgv_vec_arg(v1), tcgv_vec_arg(v2)); > + break; tcg/aarch64 seems to set TCG_TARGET_HAS_shv_vec and only do the right-shifts via expand_op. Is there a difference between the two that means Neon has to do it this way, or is it just a "works either way" thing? > + Reviewed-by: Peter Maydell <peter.mayd...@linaro.org> thanks -- PMM