In https://gcc.gnu.org/ml/gcc-patches/2019-12/msg01530.html, Segher asked me to
do the gcc_asserts as early as possible.
This patch makes sure the base register temporary is not used in the other
arguments.
I have built and bootstrapped a compiler on a little endian power8 system, and
there were no regressions in the test. In addition, I compiled both Spec 2006
and Spec 2017 benchmarks with this compiler and I saw new build failures. Can
I check this into the trunk?
2020-01-09 Michael Meissner <meiss...@linux.ibm.com>
* config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
gcc_asserts.
--- /tmp/kXfaUP_rs6000.c 2020-01-08 13:59:48.664454496 -0500
+++ gcc/config/rs6000/rs6000.c 2020-01-08 13:59:45.593410764 -0500
@@ -6772,6 +6772,9 @@ rs6000_adjust_vec_address (rtx scalar_re
rtx new_addr;
bool valid_addr_p;
+ gcc_assert (!reg_mentioned_p (base_tmp, addr));
+ gcc_assert (!reg_mentioned_p (base_tmp, element));
+
/* Vector addresses should not have PRE_INC, PRE_DEC, or PRE_MODIFY. */
gcc_assert (GET_RTX_CLASS (GET_CODE (addr)) != RTX_AUTOINC);
@@ -6781,6 +6784,10 @@ rs6000_adjust_vec_address (rtx scalar_re
element_offset = GEN_INT (INTVAL (element) * scalar_size);
else
{
+ /* All insns should use the 'Q' constraint (address is a single register)
+ if the element number is not a constant. */
+ gcc_assert (REG_P (addr) || SUBREG_P (addr));
+
int byte_shift = exact_log2 (scalar_size);
gcc_assert (byte_shift >= 0);
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: meiss...@linux.ibm.com, phone: +1 (978) 899-4797
