The patterns used by aarch64_split_sve_subreg_move only support
integer modes, so if the widest mode is a float, we should get
its integer equivalent.

Fixes gcc.target/aarch64/sel_3.c for big-endian targets.

Tested on aarch64-linux-gnu and aarch64_be-none-elf.

Richard


2020-01-16  Richard Sandiford  <richard.sandif...@arm.com>

gcc/
        * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
        aarch64_sve_int_mode to each mode.
---
 gcc/config/aarch64/aarch64.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
index ac89cc1f9c9..600a238c1f4 100644
--- a/gcc/config/aarch64/aarch64.c
+++ b/gcc/config/aarch64/aarch64.c
@@ -4827,8 +4827,8 @@ aarch64_split_sve_subreg_move (rtx dest, rtx ptrue, rtx 
src)
   /* Decide which REV operation we need.  The mode with wider elements
      determines the mode of the operands and the mode with the narrower
      elements determines the reverse width.  */
-  machine_mode mode_with_wider_elts = GET_MODE (dest);
-  machine_mode mode_with_narrower_elts = GET_MODE (src);
+  machine_mode mode_with_wider_elts = aarch64_sve_int_mode (GET_MODE (dest));
+  machine_mode mode_with_narrower_elts = aarch64_sve_int_mode (GET_MODE (src));
   if (GET_MODE_UNIT_SIZE (mode_with_wider_elts)
       < GET_MODE_UNIT_SIZE (mode_with_narrower_elts))
     std::swap (mode_with_wider_elts, mode_with_narrower_elts);

Reply via email to