aarch64-simd.md (aarch64_dup_lane<mode>): Add r<-w alternative and update.

	gcc/
	* config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>): Add r<-w
	alternative and update.
	(aarch64_dup_lanedi): Delete.
	* config/aarch64/arm_neon.h (vdup<bhsd>_lane_*): Update.
	* config/aarch64/aarch64-simd-builtins.def: Update.

	testsuite/
	* gcc.target/aarch64/scalar_intrinsics.c: Update.

From-SVN: r200152

6 files changed, 34 insertions, 33 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index a58dd00..d9e6187 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2013-06-17  Sofiane Naci  <sofiane.naci@arm.com>
+
+	* config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>): Add r<-w
+	alternative and update.
+	(aarch64_dup_lanedi): Delete.
+	* config/aarch64/arm_neon.h (vdup<bhsd>_lane_*): Update.
+	* config/aarch64/aarch64-simd-builtins.def: Update.
+
 2013-06-17  Richard Biener  <rguenther@suse.de>
 
 	* lto-streamer.h (enum LTO_tags): Add LTO_tree_scc.
diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 5134f96..4d9b966 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -64,7 +64,7 @@
   BUILTIN_VQ (REINTERP, reinterpretv2df, 0)
 
   BUILTIN_VDQ_I (BINOP, dup_lane, 0)
-  BUILTIN_SDQ_I (BINOP, dup_lane, 0)
+  BUILTIN_VDQ_I (BINOP, dup_lane_scalar, 0)
   /* Implemented by aarch64_<sur>q<r>shl<mode>.  */
   BUILTIN_VSDQ_I (BINOP, sqshl, 0)
   BUILTIN_VSDQ_I (BINOP, uqshl, 0)
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index 02037f3..08826b5 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -357,29 +357,18 @@
    (set_attr "simd_mode" "<MODE>")]
 )
 
-(define_insn "aarch64_dup_lane<mode>"
-  [(set (match_operand:ALLX 0 "register_operand" "=w")
+(define_insn "aarch64_dup_lane_scalar<mode>"
+  [(set (match_operand:<VEL> 0 "register_operand" "=w, r")
 	(vec_select:<VEL>
-	  (match_operand:<VCON> 1 "register_operand" "w")
-	  (parallel [(match_operand:SI 2 "immediate_operand" "i")])
+	  (match_operand:VDQ 1 "register_operand" "w, w")
+	  (parallel [(match_operand:SI 2 "immediate_operand" "i, i")])
         ))]
   "TARGET_SIMD"
-  "dup\\t%<v>0<Vmtype>, %1.<Vetype>[%2]"
-  [(set_attr "simd_type" "simd_dup")
-   (set_attr "simd_mode" "<MODE>")]
-)
-
-(define_insn "aarch64_dup_lanedi"
-  [(set (match_operand:DI 0 "register_operand" "=w,r")
-	(vec_select:DI
-	  (match_operand:V2DI 1 "register_operand" "w,w")
-	  (parallel [(match_operand:SI 2 "immediate_operand" "i,i")])))]
-  "TARGET_SIMD"
   "@
-   dup\\t%<v>0<Vmtype>, %1.<Vetype>[%2]
-   umov\t%0, %1.d[%2]"
-  [(set_attr "simd_type" "simd_dup")
-   (set_attr "simd_mode" "DI")]
+   dup\\t%<Vetype>0, %1.<Vetype>[%2]
+   umov\\t%<vw>0, %1.<Vetype>[%2]"
+  [(set_attr "simd_type" "simd_dup, simd_movgp")
+   (set_attr "simd_mode" "<MODE>")]
 )
 
 (define_insn "aarch64_simd_dup<mode>"
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 608db35..760ba3d 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -20234,49 +20234,49 @@ vcvtpq_u64_f64 (float64x2_t __a)
 __extension__ static __inline int8x1_t __attribute__ ((__always_inline__))
 vdupb_lane_s8 (int8x16_t a, int const b)
 {
-  return __builtin_aarch64_dup_laneqi (a, b);
+  return __builtin_aarch64_dup_lane_scalarv16qi (a, b);
 }
 
 __extension__ static __inline uint8x1_t __attribute__ ((__always_inline__))
 vdupb_lane_u8 (uint8x16_t a, int const b)
 {
-  return (uint8x1_t) __builtin_aarch64_dup_laneqi ((int8x16_t) a, b);
+  return (uint8x1_t) __builtin_aarch64_dup_lane_scalarv16qi ((int8x16_t) a, b);
 }
 
 __extension__ static __inline int16x1_t __attribute__ ((__always_inline__))
 vduph_lane_s16 (int16x8_t a, int const b)
 {
-  return __builtin_aarch64_dup_lanehi (a, b);
+  return __builtin_aarch64_dup_lane_scalarv8hi (a, b);
 }
 
 __extension__ static __inline uint16x1_t __attribute__ ((__always_inline__))
 vduph_lane_u16 (uint16x8_t a, int const b)
 {
-  return (uint16x1_t) __builtin_aarch64_dup_lanehi ((int16x8_t) a, b);
+  return (uint16x1_t) __builtin_aarch64_dup_lane_scalarv8hi ((int16x8_t) a, b);
 }
 
 __extension__ static __inline int32x1_t __attribute__ ((__always_inline__))
 vdups_lane_s32 (int32x4_t a, int const b)
 {
-  return __builtin_aarch64_dup_lanesi (a, b);
+  return __builtin_aarch64_dup_lane_scalarv4si (a, b);
 }
 
 __extension__ static __inline uint32x1_t __attribute__ ((__always_inline__))
 vdups_lane_u32 (uint32x4_t a, int const b)
 {
-  return (uint32x1_t) __builtin_aarch64_dup_lanesi ((int32x4_t) a, b);
+  return (uint32x1_t) __builtin_aarch64_dup_lane_scalarv4si ((int32x4_t) a, b);
 }
 
 __extension__ static __inline int64x1_t __attribute__ ((__always_inline__))
 vdupd_lane_s64 (int64x2_t a, int const b)
 {
-  return __builtin_aarch64_dup_lanedi (a, b);
+  return __builtin_aarch64_dup_lane_scalarv2di (a, b);
 }
 
 __extension__ static __inline uint64x1_t __attribute__ ((__always_inline__))
 vdupd_lane_u64 (uint64x2_t a, int const b)
 {
-  return (uint64x1_t) __builtin_aarch64_dup_lanedi ((int64x2_t) a, b);
+  return (uint64x1_t) __builtin_aarch64_dup_lane_scalarv2di ((int64x2_t) a, b);
 }
 
 /* vldn */
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 147ae19..cb05b1c 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,7 @@
+2013-06-17  Sofiane Naci  <sofiane.naci@arm.com>
+
+	* gcc.target/aarch64/scalar_intrinsics.c: Update.
+
 2013-06-17  Paolo Carlini  <paolo.carlini@oracle.com>
 
 	PR c++/16128
diff --git a/gcc/testsuite/gcc.target/aarch64/scalar_intrinsics.c b/gcc/testsuite/gcc.target/aarch64/scalar_intrinsics.c
index 16537ce..46d3449 100644
--- a/gcc/testsuite/gcc.target/aarch64/scalar_intrinsics.c
+++ b/gcc/testsuite/gcc.target/aarch64/scalar_intrinsics.c
@@ -1,5 +1,5 @@
 /* { dg-do compile } */
-/* { dg-options "-O2" } */
+/* { dg-options "-O2 -dp" } */
 
 #include <arm_neon.h>
 
@@ -181,7 +181,7 @@ test_vcltzd_s64 (int64x1_t a)
   return res;
 }
 
-/* { dg-final { scan-assembler-times "\\tdup\\tb\[0-9\]+, v\[0-9\]+\.b" 2 } } */
+/* { dg-final { scan-assembler-times "aarch64_dup_lane_scalarv16qi"  2 } } */
 
 int8x1_t
 test_vdupb_lane_s8 (int8x16_t a)
@@ -195,7 +195,7 @@ test_vdupb_lane_u8 (uint8x16_t a)
   return vdupb_lane_u8 (a, 2);
 }
 
-/* { dg-final { scan-assembler-times "\\tdup\\th\[0-9\]+, v\[0-9\]+\.h" 2 } } */
+/* { dg-final { scan-assembler-times "aarch64_dup_lane_scalarv8hi"  2 } } */
 
 int16x1_t
 test_vduph_lane_s16 (int16x8_t a)
@@ -209,7 +209,7 @@ test_vduph_lane_u16 (uint16x8_t a)
   return vduph_lane_u16 (a, 2);
 }
 
-/* { dg-final { scan-assembler-times "\\tdup\\ts\[0-9\]+, v\[0-9\]+\.s" 2 } } */
+/* { dg-final { scan-assembler-times "aarch64_dup_lane_scalarv4si"  2 } } */
 
 int32x1_t
 test_vdups_lane_s32 (int32x4_t a)
@@ -223,7 +223,7 @@ test_vdups_lane_u32 (uint32x4_t a)
   return vdups_lane_u32 (a, 2);
 }
 
-/* { dg-final { scan-assembler-times "\\tumov\\tx\[0-9\]+, v\[0-9\]+\.d" 2 } } */
+/* { dg-final { scan-assembler-times "aarch64_dup_lane_scalarv2di"  2 } } */
 
 int64x1_t
 test_vdupd_lane_s64 (int64x2_t a)