Canonicalize vec_merge in simplify_ternary_operation

Similar to the canonicalization done in combine, we canonicalize vec_merge with
swap_communattive_operands_p in simplify_ternary_operation too.

gcc/ChangeLog:

	* config/aarch64/aarch64-protos.h (aarch64_exact_log2_inverse): New.
	* config/aarch64/aarch64-simd.md (aarch64_simd_vec_set_zero<mode>):
	Update pattern accordingly.
	* config/aarch64/aarch64.cc (aarch64_exact_log2_inverse): New.
	* simplify-rtx.cc (simplify_context::simplify_ternary_operation):
	Canonicalize vec_merge.

Signed-off-by: Pengxuan Zheng <quic_pzheng@quicinc.com>

4 files changed, 24 insertions, 4 deletions

diff --git a/gcc/config/aarch64/aarch64-protos.h b/gcc/config/aarch64/aarch64-protos.h
index c83c35c..c935e7b 100644
--- a/gcc/config/aarch64/aarch64-protos.h
+++ b/gcc/config/aarch64/aarch64-protos.h
@@ -1055,6 +1055,7 @@ void aarch64_subvti_scratch_regs (rtx, rtx, rtx *,
 				  rtx *, rtx *, rtx *);
 void aarch64_expand_subvti (rtx, rtx, rtx,
 			    rtx, rtx, rtx, rtx, bool);
+int aarch64_exact_log2_inverse (unsigned int, rtx);
 
 
 /* Initialize builtins for SIMD intrinsics.  */
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index e2afe87..1099e74 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -1193,12 +1193,14 @@
 (define_insn "aarch64_simd_vec_set_zero<mode>"
   [(set (match_operand:VALL_F16 0 "register_operand" "=w")
 	(vec_merge:VALL_F16
-	    (match_operand:VALL_F16 1 "aarch64_simd_imm_zero" "")
-	    (match_operand:VALL_F16 3 "register_operand" "0")
+	    (match_operand:VALL_F16 1 "register_operand" "0")
+	    (match_operand:VALL_F16 3 "aarch64_simd_imm_zero" "")
 	    (match_operand:SI 2 "immediate_operand" "i")))]
-  "TARGET_SIMD && exact_log2 (INTVAL (operands[2])) >= 0"
+  "TARGET_SIMD && aarch64_exact_log2_inverse (<nunits>, operands[2]) >= 0"
   {
-    int elt = ENDIAN_LANE_N (<nunits>, exact_log2 (INTVAL (operands[2])));
+    int elt = ENDIAN_LANE_N (<nunits>,
+			     aarch64_exact_log2_inverse (<nunits>,
+							 operands[2]));
     operands[2] = GEN_INT ((HOST_WIDE_INT) 1 << elt);
     return "ins\\t%0.<Vetype>[%p2], <vwcore>zr";
   }
diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
index 2dc5f4c..9e3f288 100644
--- a/gcc/config/aarch64/aarch64.cc
+++ b/gcc/config/aarch64/aarch64.cc
@@ -23914,6 +23914,16 @@ aarch64_strided_registers_p (rtx *operands, unsigned int num_operands,
   return true;
 }
 
+/* Return the base 2 logarithm of the bit inverse of OP masked by the lowest
+   NELTS bits, if OP is a power of 2.  Otherwise, returns -1.  */
+
+int
+aarch64_exact_log2_inverse (unsigned int nelts, rtx op)
+{
+  return exact_log2 ((~INTVAL (op))
+		     & ((HOST_WIDE_INT_1U << nelts) - 1));
+}
+
 /* Bounds-check lanes.  Ensure OPERAND lies between LOW (inclusive) and
    HIGH (exclusive).  */
 void
diff --git a/gcc/simplify-rtx.cc b/gcc/simplify-rtx.cc
index 7bcbe11..b34fd2f 100644
--- a/gcc/simplify-rtx.cc
+++ b/gcc/simplify-rtx.cc
@@ -7387,6 +7387,13 @@ simplify_context::simplify_ternary_operation (rtx_code code, machine_mode mode,
 	      return gen_rtx_CONST_VECTOR (mode, v);
 	    }
 
+	  if (swap_commutative_operands_p (op0, op1)
+	      /* Two operands have same precedence, then first bit of mask
+		 select first operand.  */
+	      || (!swap_commutative_operands_p (op1, op0) && !(sel & 1)))
+	    return simplify_gen_ternary (code, mode, mode, op1, op0,
+					 GEN_INT (~sel & mask));
+
 	  /* Replace (vec_merge (vec_merge a b m) c n) with (vec_merge b c n)
 	     if no element from a appears in the result.  */
 	  if (GET_CODE (op0) == VEC_MERGE)