[PATCH] dse: Use SUBREG_REG for copy_to_mode_reg in DSE replace_read for WORD_REGISTER_OPERATIONS targets [PR109040]

[Jakub Jelinek via Gcc-patches]

Hi!

While we've agreed this is not the right fix for the PR109040 bug,
the patch clearly improves generated code (at least on the testcase from the
PR), so I'd like to propose this as optimization heuristics improvement
for GCC 14.

Ok for trunk?

2023-04-18  Jakub Jelinek  <ja...@redhat.com>

        PR target/109040
        * dse.cc (replace_read): If read_reg is a SUBREG of a word mode
        REG, for WORD_REGISTER_OPERATIONS copy SUBREG_REG of it into
        a new REG rather than the SUBREG.

--- gcc/dse.cc.jj       2023-01-02 09:32:50.369880943 +0100
+++ gcc/dse.cc  2023-04-04 22:17:22.906347794 +0200
@@ -2012,7 +2012,19 @@ replace_read (store_info *store_info, in
     }
   /* Force the value into a new register so that it won't be clobbered
      between the store and the load.  */
-  read_reg = copy_to_mode_reg (read_mode, read_reg);
+  if (WORD_REGISTER_OPERATIONS
+      && GET_CODE (read_reg) == SUBREG
+      && REG_P (SUBREG_REG (read_reg))
+      && GET_MODE (SUBREG_REG (read_reg)) == word_mode)
+    {
+      /* For WORD_REGISTER_OPERATIONS with subreg of word_mode register
+        force SUBREG_REG into a new register rather than the SUBREG.  */
+      rtx r = copy_to_mode_reg (word_mode, SUBREG_REG (read_reg));
+      read_reg = shallow_copy_rtx (read_reg);
+      SUBREG_REG (read_reg) = r;
+    }
+  else
+    read_reg = copy_to_mode_reg (read_mode, read_reg);
   insns = get_insns ();
   end_sequence ();
 

        Jakub