[AArch64] Optimize prologue when there is no frame pointer.

From-SVN: r212999

gcc/ChangeLog

@@ -1,3 +1,8 @@
+2014-07-24  Jiong Wang  <jiong.wang@arm.com>
+
+	* config/aarch64/aarch64.c (aarch64_pushwb_single_reg): New function.
+	(aarch64_expand_prologue): Optimize prologue when !frame_pointer_needed.
+
 2014-07-24  Jiong Wang  <jiong.wang@arm.com>
 
 	* config/aarch64/aarch64.c (aarch64_restore_callee_saves)

gcc/config/aarch64/aarch64.c

@@ -1930,6 +1930,22 @@ aarch64_next_callee_save (unsigned regno, unsigned limit)
   return regno;
 }
 
+static void
+aarch64_pushwb_single_reg (enum machine_mode mode, unsigned regno,
+			   HOST_WIDE_INT adjustment)
+ {
+  rtx base_rtx = stack_pointer_rtx;
+  rtx insn, reg, mem;
+
+  reg = gen_rtx_REG (mode, regno);
+  mem = gen_rtx_PRE_MODIFY (Pmode, base_rtx,
+			    plus_constant (Pmode, base_rtx, -adjustment));
+  mem = gen_rtx_MEM (mode, mem);
+
+  insn = emit_move_insn (mem, reg);
+  RTX_FRAME_RELATED_P (insn) = 1;
+}
+
 static rtx
 aarch64_gen_storewb_pair (enum machine_mode mode, rtx base, rtx reg, rtx reg2,
 			  HOST_WIDE_INT adjustment)
@@ -2279,11 +2295,10 @@ aarch64_expand_prologue (void)
     {
       bool skip_wb = false;
 
-      /* Save the frame pointer and lr if the frame pointer is needed
-	 first.  Make the frame pointer point to the location of the
-	 old frame pointer on the stack.  */
       if (frame_pointer_needed)
 	{
+	  skip_wb = true;
+
 	  if (fp_offset)
 	    {
 	      insn = emit_insn (gen_add2_insn (stack_pointer_rtx,
@@ -2291,12 +2306,11 @@ aarch64_expand_prologue (void)
 	      RTX_FRAME_RELATED_P (insn) = 1;
 	      aarch64_set_frame_expr (gen_rtx_SET
 				      (Pmode, stack_pointer_rtx,
-				       gen_rtx_MINUS (Pmode,
+				       gen_rtx_MINUS (Pmode, stack_pointer_rtx,
-						      stack_pointer_rtx,
 						      GEN_INT (offset))));
 
 	      aarch64_save_callee_saves (DImode, fp_offset, R29_REGNUM,
-					 R30_REGNUM, skip_wb);
+					 R30_REGNUM, false);
 	    }
 	  else
 	    aarch64_pushwb_pair_reg (DImode, R29_REGNUM, R30_REGNUM, offset);
@@ -2314,20 +2328,36 @@ aarch64_expand_prologue (void)
 	  RTX_FRAME_RELATED_P (insn) = 1;
 	  insn = emit_insn (gen_stack_tie (stack_pointer_rtx,
 					   hard_frame_pointer_rtx));
-
-	  aarch64_save_callee_saves (DImode, fp_offset, R0_REGNUM, R28_REGNUM,
-				     skip_wb);
 	}
       else
 	{
-	  insn = emit_insn (gen_add2_insn (stack_pointer_rtx,
+	  unsigned reg1 = cfun->machine->frame.wb_candidate1;
-					   GEN_INT (-offset)));
+	  unsigned reg2 = cfun->machine->frame.wb_candidate2;
-	  RTX_FRAME_RELATED_P (insn) = 1;
 
-	  aarch64_save_callee_saves (DImode, fp_offset, R0_REGNUM, R30_REGNUM,
+	  if (fp_offset
-				     skip_wb);
+	      || reg1 == FIRST_PSEUDO_REGISTER
+	      || (reg2 == FIRST_PSEUDO_REGISTER
+		  && offset >= 256))
+	    {
+	      insn = emit_insn (gen_add2_insn (stack_pointer_rtx,
+					       GEN_INT (-offset)));
+	      RTX_FRAME_RELATED_P (insn) = 1;
+	    }
+	  else
+	    {
+	      enum machine_mode mode1 = (reg1 <= R30_REGNUM) ? DImode : DFmode;
+
+	      skip_wb = true;
+
+	      if (reg2 == FIRST_PSEUDO_REGISTER)
+		aarch64_pushwb_single_reg (mode1, reg1, offset);
+	      else
+		aarch64_pushwb_pair_reg (mode1, reg1, reg2, offset);
+	    }
 	}
 
+      aarch64_save_callee_saves (DImode, fp_offset, R0_REGNUM, R30_REGNUM,
+				 skip_wb);
       aarch64_save_callee_saves (DFmode, fp_offset, V0_REGNUM, V31_REGNUM,
 				 skip_wb);
     }

gcc/testsuite/ChangeLog

@@ -1,3 +1,15 @@
+2014-07-24  Jiong Wang  <jiong.wang@arm.com>
+
+	* gcc.target/aarch64/test_frame_1.c: Match optimized instruction
+	sequences.
+	* gcc.target/aarch64/test_frame_10.c: Likewise.
+	* gcc.target/aarch64/test_frame_2.c: Likewise.
+	* gcc.target/aarch64/test_frame_4.c: Likewise.
+	* gcc.target/aarch64/test_frame_6.c: Likewise.
+	* gcc.target/aarch64/test_frame_7.c: Likewise.
+	* gcc.target/aarch64/test_frame_8.c: Likewise.
+	* gcc.target/aarch64/test_fp_attribute_1.c: Likewise.
+
 2014-07-24  Martin Jambor  <mjambor@suse.cz>
 
 	PR ipa/61160

gcc/testsuite/gcc.target/aarch64/test_fp_attribute_1.c

@@ -21,6 +21,6 @@ non_leaf_2 (void)
   leaf ();
 }
 
-/* { dg-final { scan-assembler-times "str\tx30, \\\[sp\\\]" 2 } } */
+/* { dg-final { scan-assembler-times "str\tx30, \\\[sp, -\[0-9\]+\\\]!" 2 } } */
 
 /* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_1.c

@@ -6,9 +6,12 @@
      * optimized code should use "str !" for stack adjustment.  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern (test1, 200, )
 t_frame_run (test1)
+
+/* { dg-final { scan-assembler-times "str\tx30, \\\[sp, -\[0-9\]+\\\]!" 2 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_10.c

@@ -8,9 +8,12 @@
        the first subtractions could be optimized into "stp !".  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern_outgoing (test10, 480, "x19", 24, a[8], a[9], a[10])
 t_frame_run (test10)
+
+/* { dg-final { scan-assembler-times "stp\tx19, x30, \\\[sp, -\[0-9\]+\\\]!" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_2.c

@@ -6,9 +6,13 @@
      * optimized code should use "stp !" for stack adjustment.  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern (test2, 200, "x19")
 t_frame_run (test2)
+
+
+/* { dg-final { scan-assembler-times "stp\tx19, x30, \\\[sp, -\[0-9\]+\\\]!" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_4.c

@@ -6,9 +6,12 @@
      * we can use "stp !" to optimize stack adjustment.  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern (test4, 400, "x19")
 t_frame_run (test4)
+
+/* { dg-final { scan-assembler-times "stp\tx19, x30, \\\[sp, -\[0-9\]+\\\]!" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_6.c

@@ -7,9 +7,12 @@
        the second subtraction should use "str !".  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern (test6, 700, )
 t_frame_run (test6)
+
+/* { dg-final { scan-assembler-times "str\tx30, \\\[sp, -\[0-9\]+\\\]!" 2 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_7.c

@@ -7,9 +7,12 @@
        the second subtraction should use "stp !".  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern (test7, 700, "x19")
 t_frame_run (test7)
+
+/* { dg-final { scan-assembler-times "stp\tx19, x30, \\\[sp, -\[0-9\]+\\\]!" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/test_frame_8.c

@@ -5,9 +5,12 @@
      * number of callee-saved reg == 1.  */
 
 /* { dg-do run } */
-/* { dg-options "-O2 -fomit-frame-pointer" } */
+/* { dg-options "-O2 -fomit-frame-pointer --save-temps" } */
 
 #include "test_frame_common.h"
 
 t_frame_pattern_outgoing (test8, 700, , 8, a[8])
 t_frame_run (test8)
+
+/* { dg-final { scan-assembler-times "str\tx30, \\\[sp, -\[0-9\]+\\\]!" 3 } } */
+/* { dg-final { cleanup-saved-temps } } */