[AArch64][4/5] Implement fusion of ARDP+LDR

	* config/aarch64/aarch64.c (AARCH64_FUSE_ADRP_LDR): Define.
	(cortexa53_tunings): Specify AARCH64_FUSE_ADRP_LDR in fuseable_ops.
	(aarch_macro_fusion_pair_p): Handle AARCH64_FUSE_ADRP_LDR.

From-SVN: r218014

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index d58c8c3593b288d86932cd197dab4c468d4281ea..60a6b22a8942c2e25ffb4b7d2047dd3f0dd0ffcd 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,9 @@
+2014-11-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
+
+       * config/aarch64/aarch64.c (AARCH64_FUSE_ADRP_LDR): Define.
+       (cortexa53_tunings): Specify AARCH64_FUSE_ADRP_LDR in fuseable_ops.
+       (aarch_macro_fusion_pair_p): Handle AARCH64_FUSE_ADRP_LDR.
+
 2014-11-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 
        * config/aarch64/aarch64.c (AARCH64_FUSE_MOVK_MOVK): Define.

diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
index 5598b5b253edaeca100c98948a36068a5c66bf9a..ee9a9624f16c1ef9a308a050c76c3153ba2fd0bb 100644 (file)
--- a/gcc/config/aarch64/aarch64.c
+++ b/gcc/config/aarch64/aarch64.c
@@ -309,6 +309,7 @@ static const struct cpu_vector_cost cortexa57_vector_cost =
 #define AARCH64_FUSE_MOV_MOVK  (1 << 0)
 #define AARCH64_FUSE_ADRP_ADD  (1 << 1)
 #define AARCH64_FUSE_MOVK_MOVK (1 << 2)
+#define AARCH64_FUSE_ADRP_LDR  (1 << 3)
 
 #if HAVE_DESIGNATED_INITIALIZERS && GCC_VERSION >= 2007
 __extension__
@@ -332,7 +333,8 @@ static const struct tune_params cortexa53_tunings =
   &generic_vector_cost,
   NAMED_PARAM (memmov_cost, 4),
   NAMED_PARAM (issue_rate, 2),
-  NAMED_PARAM (fuseable_ops, (AARCH64_FUSE_MOV_MOVK | AARCH64_FUSE_ADRP_ADD | AARCH64_FUSE_MOVK_MOVK))
+  NAMED_PARAM (fuseable_ops, (AARCH64_FUSE_MOV_MOVK | AARCH64_FUSE_ADRP_ADD
+                             | AARCH64_FUSE_MOVK_MOVK | AARCH64_FUSE_ADRP_LDR))
 };
 
 static const struct tune_params cortexa57_tunings =
@@ -10485,6 +10487,37 @@ aarch_macro_fusion_pair_p (rtx_insn *prev, rtx_insn *curr)
         return true;
 
     }
+  if (simple_sets_p
+      && (aarch64_tune_params->fuseable_ops & AARCH64_FUSE_ADRP_LDR))
+    {
+      /* We're trying to match:
+          prev (adrp) == (set (reg r0)
+                              (high (symbol_ref ("SYM"))))
+          curr (ldr) == (set (reg r1)
+                             (mem (lo_sum (reg r0)
+                                             (symbol_ref ("SYM")))))
+                 or
+          curr (ldr) == (set (reg r1)
+                             (zero_extend (mem
+                                           (lo_sum (reg r0)
+                                                   (symbol_ref ("SYM"))))))  */
+      if (satisfies_constraint_Ush (SET_SRC (prev_set))
+          && REG_P (SET_DEST (prev_set)) && REG_P (SET_DEST (curr_set)))
+        {
+          rtx curr_src = SET_SRC (curr_set);
+
+          if (GET_CODE (curr_src) == ZERO_EXTEND)
+            curr_src = XEXP (curr_src, 0);
+
+          if (MEM_P (curr_src) && GET_CODE (XEXP (curr_src, 0)) == LO_SUM
+              && REG_P (XEXP (XEXP (curr_src, 0), 0))
+              && REGNO (XEXP (XEXP (curr_src, 0), 0))
+                 == REGNO (SET_DEST (prev_set))
+              && rtx_equal_p (XEXP (XEXP (curr_src, 0), 1),
+                              XEXP (SET_SRC (prev_set), 0)))
+              return true;
+        }
+    }
 
   return false;
 }