tcg: Fix tcg gen for vectorized absolute value

The fallback inline expansion for vectorized absolute value,
when the host doesn't support such an insn was flawed.

E.g. when a vector of bytes has all elements negative, mask
will be 0xffff_ffff_ffff_ffff. Subtracting mask only adds 1
to the low element instead of all elements becase -mask is 1
and not 0x0101_0101_0101_0101.

Backports commit e7e8f33fb603c3bfa0479d7d924f2ad676a84317
This commit is contained in:
Stephen Long 2021-03-01 18:04:37 -05:00 committed by Lioncash
parent cefb1666c0
commit c9dc750058

View file

@ -2270,7 +2270,8 @@ static void gen_absv_mask(TCGContext *s, TCGv_i64 d, TCGv_i64 b, unsigned vece)
* so we never have carry into the next element.
*/
tcg_gen_xor_i64(s, d, b, t);
tcg_gen_sub_i64(s, d, d, t);
tcg_gen_andi_i64(s, t, t, dup_const(vece, 1));
tcg_gen_add_i64(s, d, d, t);
tcg_temp_free_i64(s, t);
}