@@ -1041,7 +1041,7 @@ static void gen_ushl_vec(unsigned vece, TCGv_vec dst,
TCGv_vec rval = tcg_temp_new_vec_matching(dst);
TCGv_vec lsh = tcg_temp_new_vec_matching(dst);
TCGv_vec rsh = tcg_temp_new_vec_matching(dst);
- TCGv_vec max;
+ TCGv_vec max, zero;
tcg_gen_neg_vec(vece, rsh, shift);
if (vece == MO_8) {
@@ -1061,23 +1061,20 @@ static void gen_ushl_vec(unsigned vece, TCGv_vec dst,
tcg_gen_shrv_vec(vece, rval, src, rsh);
/*
- * The choice of LT (signed) and GEU (unsigned) are biased toward
+ * The choice of GE (signed) and GEU (unsigned) are biased toward
* the instructions of the x86_64 host. For MO_8, the whole byte
* is significant so we must use an unsigned compare; otherwise we
* have already masked to a byte and so a signed compare works.
* Other tcg hosts have a full set of comparisons and do not care.
*/
+ zero = tcg_constant_vec_matching(dst, vece, 0);
max = tcg_constant_vec_matching(dst, vece, 8 << vece);
if (vece == MO_8) {
- tcg_gen_cmp_vec(TCG_COND_GEU, vece, lsh, lsh, max);
- tcg_gen_cmp_vec(TCG_COND_GEU, vece, rsh, rsh, max);
- tcg_gen_andc_vec(vece, lval, lval, lsh);
- tcg_gen_andc_vec(vece, rval, rval, rsh);
+ tcg_gen_cmpsel_vec(TCG_COND_GEU, vece, lval, lsh, max, zero, lval);
+ tcg_gen_cmpsel_vec(TCG_COND_GEU, vece, rval, rsh, max, zero, rval);
} else {
- tcg_gen_cmp_vec(TCG_COND_LT, vece, lsh, lsh, max);
- tcg_gen_cmp_vec(TCG_COND_LT, vece, rsh, rsh, max);
- tcg_gen_and_vec(vece, lval, lval, lsh);
- tcg_gen_and_vec(vece, rval, rval, rsh);
+ tcg_gen_cmpsel_vec(TCG_COND_GE, vece, lval, lsh, max, zero, lval);
+ tcg_gen_cmpsel_vec(TCG_COND_GE, vece, rval, rsh, max, zero, rval);
}
tcg_gen_or_vec(vece, dst, lval, rval);
}
@@ -1087,7 +1084,7 @@ void gen_gvec_ushl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
{
static const TCGOpcode vecop_list[] = {
INDEX_op_neg_vec, INDEX_op_shlv_vec,
- INDEX_op_shrv_vec, INDEX_op_cmp_vec, 0
+ INDEX_op_shrv_vec, INDEX_op_cmpsel_vec, 0
};
static const GVecGen3 ops[4] = {
{ .fniv = gen_ushl_vec,
Instead of cmp+and or cmp+andc, use cmpsel. This will be better for hosts that use predicate registers for cmp. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/arm/tcg/gengvec.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-)