xtensa: Eliminate unnecessary general-purpose reg-reg moves

Register-register move instructions that can be easily seen as
unnecessary by the human eye may remain in the compiled result.
For example:

/* example */
double test(double a, double b) {
  return __builtin_copysign(a, b);
}

test:
	add.n	a3, a3, a3
	extui	a5, a5, 31, 1
	ssai	1
				;; Be in the same BB
	src	a7, a5, a3	;; Replacing the destination doesn't
				;;   violate any constraints of the
				;;   operands
				;; No CALL insns in this span
				;; Both A3 and A7 are irrelevant to
				;;   insns in this span
	mov.n	a3, a7		;; An unnecessary reg-reg move
				;; A7 is not used after this
	ret.n

The last two instructions above, excluding the return instruction,
could be done like this:

	src	a3, a5, a3

This symptom often occurs when handling DI/DFmode values with SImode
instructions.  This patch solves the above problem using peephole2
pattern.

gcc/ChangeLog:

	* config/xtensa/xtensa.md: New peephole2 pattern that eliminates
	the occurrence of general-purpose register used only once and for
	transferring intermediate value.

gcc/testsuite/ChangeLog:

	* gcc.target/xtensa/elim_GP_regmove_0.c: New test.
	* gcc.target/xtensa/elim_GP_regmove_1.c: New test.

3 files changed, 79 insertions, 0 deletions

diff --git a/gcc/config/xtensa/xtensa.md b/gcc/config/xtensa/xtensa.md
index d6116d6..cf25beb 100644
--- a/gcc/config/xtensa/xtensa.md
+++ b/gcc/config/xtensa/xtensa.md
@@ -3050,3 +3050,49 @@ FALLTHRU:;
   operands[1] = GEN_INT (imm0);
   operands[2] = GEN_INT (imm1);
 })
+
+(define_peephole2
+  [(set (match_operand 0 "register_operand")
+	(match_operand 1 "register_operand"))]
+  "REG_NREGS (operands[0]) == 1 && GP_REG_P (REGNO (operands[0]))
+   && REG_NREGS (operands[1]) == 1 && GP_REG_P (REGNO (operands[1]))
+   && peep2_reg_dead_p (1, operands[1])"
+  [(const_int 0)]
+{
+  basic_block bb = BLOCK_FOR_INSN (curr_insn);
+  rtx_insn *head = BB_HEAD (bb), *insn;
+  rtx dest = operands[0], src = operands[1], pattern, t_dest, dest_orig;
+  for (insn = PREV_INSN (curr_insn);
+       insn && insn != head;
+       insn = PREV_INSN (insn))
+    if (CALL_P (insn))
+      break;
+    else if (INSN_P (insn))
+      {
+	if (GET_CODE (pattern = PATTERN (insn)) == SET
+	    && REG_P (t_dest = SET_DEST (pattern))
+	    && REG_NREGS (t_dest) == 1
+	    && REGNO (t_dest) == REGNO (src))
+	{
+	  dest_orig = SET_DEST (pattern);
+	  SET_DEST (pattern) = gen_rtx_REG (GET_MODE (t_dest),
+					    REGNO (dest));
+	  extract_insn (insn);
+	  if (!constrain_operands (true, get_enabled_alternatives (insn)))
+	    {
+	      SET_DEST (pattern) = dest_orig;
+	      goto ABORT;
+	    }
+	  df_insn_rescan (insn);
+	  goto FALLTHRU;
+	}
+	if (reg_overlap_mentioned_p (dest, pattern)
+	    || reg_overlap_mentioned_p (src, pattern)
+	    || set_of (dest, insn)
+	    || set_of (src, insn))
+	  break;
+      }
+ABORT:
+  FAIL;
+FALLTHRU:;
+})
diff --git a/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_0.c b/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_0.c
new file mode 100644
index 0000000..5c195c3
--- /dev/null
+++ b/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_0.c
@@ -0,0 +1,23 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fpeephole2" } */
+
+/* can be processed */
+double test0(double a, double b) {
+  return __builtin_copysign(a, b);
+}
+
+/* cannot be processed: due to violate '0' constraint of the 2nd source operand.  */
+int test1(int a, int b) {
+  int c;
+  asm volatile ("" : "=a"(c) : "r"(a), "0"(b));
+  return c;
+}
+
+/* cannot be processed: due to violate '&' constraint of the destination operand.  */
+int test2(int a) {
+  int b;
+  asm volatile ("" : "=&a"(b) : "r"(a));
+  return b;
+}
+
+/* { dg-final { scan-assembler-times "mov\t|mov.n\t" 2 } } */
diff --git a/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_1.c b/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_1.c
new file mode 100644
index 0000000..a13ef81
--- /dev/null
+++ b/gcc/testsuite/gcc.target/xtensa/elim_GP_regmove_1.c
@@ -0,0 +1,10 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fpeephole2 -mabi=windowed" } */
+
+/* cannot be processed: due to violate 'a' constraint of the destination operand of the stack adjustment instruction.  */
+void test(void) {
+  int buffer[8192];
+  asm volatile ("" : : "m"(buffer));
+}
+
+/* { dg-final { scan-assembler-times "movsp" 1 } } */